- Email: [email protected]
A fuzzy decision system for genetically modified plant environmental risk assessment using Mamdani inference
ESWA 9580
No. of Pages 7, Model 5G
15 October 2014 Expert Systems with Applications xxx (2014) xxx–xxx 1
Contents lists available at ScienceDirect
Expert Systems with Applications journal homepage: www.elsevier.com/locate/eswa 5 6
A fuzzy decision system for genetically modified plant environmental risk assessment using Mamdani inference
3 4 7
Q1
8
Francesco Camastra a,⇑, Angelo Ciaramella a, Valeria Giovannelli b, Matteo Lener b, Valentina Rastelli b, Antonino Staiano a, Giovanni Staiano b, Alfredo Starace c a
9 10 11
b c
12 13
Department of Science and Technology, University of Naples Parthenope, Centro Direzionale Isola C4, 80143 Naples, Italy Nature Protection Department, Institute for Environmental Protection and Research (ISPRA), via v. Brancati 48, 00144 Roma, Italy Oracle Ireland Ltd., Block C, East Point Business Park, Dublin 3, Ireland
a r t i c l e
1 2 5 5 16 17
i n f o
Article history: Available online xxxx
18 19 20 21 22 23 24
Keywords: Fuzzy decision system Mamdani fuzzy inference Risk assessment Genetically modified plants Fuzzy Control Language
a b s t r a c t Environmental risk assessment (ERA) of the deliberate release of genetically modified plants (GMPs) is currently performed by human experts on the basis of own personal experience and knowledge. In this paper we describe a fuzzy decision system (FDS) for the ERA of GMPs, based on Mamdani fuzzy inference. The risk assessment in the FDS is obtained by using a fuzzy inference system (FIS), performed using jFuzzyLogic library. The FDS permits obtaining an evaluation process for the identification of potential impacts that can achieve one or more receptors through a set of migration paths. The decisions derived by FDS have been validated on real world cases by the human experts that are in charge of ERA. They have confirmed the reliability and correctness of the fuzzy system decisions. Ó 2014 Elsevier Ltd. All rights reserved.
26 27 28 29 30 31 32 33 34 35 36
37 38
1. Introduction
39
The development of genetic engineering in the last years produced a very high number of genetically modified plants (GMPs). Whereas in USA the use of GMPs is widely spread in agriculture, in Europe there are discordant policies w.r.t. GMPs. For instance, commercialization of food and feed containing or consisting of GMPs is duly approved in European Community (EC), while cultivation of new genetically modified crops are not adopted. The maize MON 810, approved by the old EC legislation framework, is currently the unique GMP cultivated in the EC (e.g., Czech Republic, Poland, Spain, Portugal, Romania and Slovakia), although the evaluations about its effect on the surrounding environment are discordant (Camastra, Ciaramella, & Staiano, 2014; Lang, Brunzel, Dolek, Otto, & Thei, 2011; Perry et al., 2010). According to EC, the environmental release of GMPs is ruled by Directive 200118EC and Regulation 18292003EC. The Directive refers to the deliberate release into the environment of GMPs and sets out two regulatory regimes: Part C for the placing on the market and Part B for the deliberate release for any other purpose, i.e., field
40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56
Q2
⇑ Corresponding author. E-mail addresses: [email protected] (F. Camastra), angelo.ciaramella@uni parthenope.it (A. Ciaramella), [email protected] (V. Giovannelli), [email protected] (M. Lener), [email protected] (V. Rastelli), [email protected] (A. Staiano), giovanni.staiano@ isprambiente.it (G. Staiano), [email protected] (A. Starace).
trials (Sorlini et al., 2003). In both legislations the notifier, i.e., the person who requests the release into the environment of GMP, must perform an environmental risk assessment (ERA) on the issue. The ERA is formally defined as ‘‘the evaluation of risks to human health and the environment, whether direct or indirect, immediate or delayed, which the deliberate release or the placing on the market of GMPs may pose’’. ERA should be carried out case by case, meaning that its conclusion may depends on the GMPs and trait concerned, their intended uses, and the potential receiving environments. The ERA process should lead to the identification and evaluation of potential adverse effects of the GMP, and, at the same time, it should be conducted with a view for identifying if there is a need for risk management and it should provide the basis for the monitoring plans. The aim of this work is the development of a fuzzy decision system that should advise the notifier in performing the ERA about the cultivation of a specific GMP. ERA process is often performed in presence of incomplete and imprecise data. Moreover, it is generally yielded using the personal experience and knowledge of the notifier. Therefore the usage of fuzzy reasoning in the ERA is particularly appropriate as witnessed by the extensive application of fuzzy reasoning to the risk assessment in disparate fields (Bukhari, Tusseyeva, Lee, & Kim, 2013; Chen & Weng, 2009; Chen, Zhao, & Lee, 2010; Cho, Choi, & Kim, 2002; Davidson, Ryks, & Fazil, 2006; Guimara & Lapa, 2007; Kahraman & Kaya, 2009; Karimi & Hullermeier, 2007; Li, Zhou, Xie, & Xiong, 2008;
http://dx.doi.org/10.1016/j.eswa.2014.09.041 0957-4174/Ó 2014 Elsevier Ltd. All rights reserved.
Please cite this article in press as: Camastra, F., et al. A fuzzy decision system for genetically modified plant environmental risk assessment using Mamdani inference. Expert Systems with Applications (2014), http://dx.doi.org/10.1016/j.eswa.2014.09.041
57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82
ESWA 9580
No. of Pages 7, Model 5G
15 October 2014 2 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99
F. Camastra et al. / Expert Systems with Applications xxx (2014) xxx–xxx
Mokhtari, Ren, Roberts, & Wang, 2012; Ngai & Wat, 2003; Petrovic et al., 2014; Sadiq & Husain, 2005; Samantra, Datta, & Shankar Manapantra, 2014; Smith & Eloff, 2000; Wang & Elhag, 2008; Yang, Bonsall, & Wang, 2008). To our best knowledge, no Fuzzybased Systems for ERA of GMPs have been developed, yet. For this reason, the fuzzy decision system (FDS), object of the paper, represents a novelty. The FDS is inspired by Sorlini et al.’s methodological proposal of performing ERA on GMP field trials (Sorlini et al., 2003). The methodology allows describing the relationships between potential receptors and the harmful characteristics of a GMP field trial, leading to the identification of potential impacts. The paper is organized as follows: In Section 2 a brief overview of works, that apply fuzzy reasoning to the risk assessment, is provided; In Section 3 the methodological proposal that inspires the system is presented; The FDS structure is discussed in Section 4; Section 5 describes how the system validation has been performed; Finally some conclusions are drawn in Section 6.
100
2. Fuzzy based environmental risk assessment
101
In these last years several works have been conducted for the risk assessment in various scientific areas. Several applications were developed for the monitoring of industrial processes and their environmental impacts. Darbra et al. introduced a fuzzy logic model to assess preliminarily the risk of accidental releases of ecotoxic substances in hazard plants (Darbra, Demichela, & Mure, 2008). The tool has been tested with a set of storage yards of ecotoxic substances, mainly oil, in the Piedmont Region areas (Italy). Huang et al. proposed an interval parameter fuzzy relation analysis model for environmental risk assessment of petroleumcontaminated aquifers due to leakage from underground storage tanks (Huang, Chen, Tontiwachwuthikul, & Chakma, 2010). The modeling results have provided bases for determining desirable site remediation actions. Sadiq and Husain developed a fuzzy based methodology for estimating aggregative risk of various environmental activities, pollution sources and routes in a given process (Sadiq & Husain, 2005). The developed methodology was applied to a case study of offshore drilling waste for evaluating various discharge scenarios. Chen et al. and Li et al. applied fuzzy and statistical mechanisms for environmental risk assessment (Li, Huang, Maqsood, & Huang, 2007), e.g., pollutant dispersion for the prediction of the environmental risks (Chen et al., 2010), petroleum-contaminated groundwater system in western Canada (Li et al., 2008). Zhou proposed a set-pair analysis based on a fuzzy assessment method for the real-time monitoring and control of major hazard installations storing flammable gas (Zhou, 2010). Bajpai adopted a security risk assessment technique for chemical process industries handling hazardous chemicals in bulk (Bajpai, Sachdeva, & Gupta, 2010). Guimara and Lapa presented a nuclear case study (Guimara & Lapa, 2007), in which a fuzzy inference system was used as alternative approach in risk analysis. A standard four-loop pressurized water reactor containment cooling system was used as example case. Elsayed developed a multiple attribute risk assessment approach using a fuzzy inference system (Elsayed, 2009) and used it as an alternative approach to qualitative risk matrix techniques. The system is currently used in many industries and ship classification societies. Zeng et al. and Dikmen et al. proposed a fuzzy risk assessment methodology for construction projects (Dikmen, Birgonul, & Han, 2007; Zeng, An, & Smith, 2007). Wang and Elhag adopted a neuro-fuzzy system for bridge risks (Wang & Elhag, 2008) evaluated periodically so that the bridges with high risks can be maintained timely. Hadjimichael developed a fuzzy based tool, valuable to airline safety departments, for examining risk trends, to pilots and dispatchers for assessing risks associated with each flight, and to airline management for quantifying the effects of making safety-related changes (Hadjimichael,
102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146
2009). Kahraman and Kaya adopted a fuzzy based system as a systematic procedure for predicting potential risks to human health or the environment (Kahraman & Kaya, 2009). The proposed methodology is applied to measure drought effects by analyzing the fullness rates of the dams in Istanbul. Davidson et al. developed a fuzzy risk assessment tool for early-stage risk assessment of microbial hazards in food systems (Davidson et al., 2006). Karimi and Hullermeier, and Chongfu presented systems for assessing the risk of natural disasters (Chongfu, 1996; Karimi & Hullermeier, 2007), particularly under highly uncertain conditions. The systems has been applied for assessing the earthquake risk in metropolitan area. Recently, fuzzy based approaches are used in risk assessment model of mining equipment failure (Petrovic et al., 2014), in a real-time multi-vessel collision risk assessment system (Bukhari et al., 2013) and in risk assessment in IT outsourcing (Samantra et al., 2014). We have also to mention that approaches for risk assessment, alternative to fuzzy reasoning, have been proposed. To this purpose, we have to remark that SVM (Harris, 2013) and hybrid neural networks for credit risk assessment (Oreski & Oreski, 2014), and the Bayesian networks for risk assessment of tunneling-induced damage (Wang, Ding, Luo, & Love, 2014) have been used. Finally, we conclude the overview quoting two works, even if not strictly related to ERA, that have influenced the development of the FDS, object of the paper. The former introduced the questionnaire data mining problem and defined the rule patterns that can be mined from questionnaire data (Chen & Weng, 2009). The latter described the development of a fuzzy expert system for hotel selection (Ngai & Wat, 2003). The viability of the system as an effective procedure for hotel selection was ascertained by the positive feedback obtained from the survey questionnaires.
147
3. The methodological approach
178
The methodological proposal, that has inspired the system is based on a conceptual model developed by ISPRA1 experts (Sorlini et al., 2003). The schema, shown in Fig. 1, illustrates the possible paths of the impact from a specific source to a given receptor through disparate diffusion factors and migration routes. The model implies that the notifier compiles an electronic questionnaire. The notifier answers are collected in a relational database management system and, in a second time, become input of a fuzzy decision engine that is the system core and provides to the notifier the overall evaluation of risk assessment related to a specific GMP. The questionnaire can be grouped in specific sets of questions where each set corresponds to a specific box of the diagram of the conceptual model. For each block the potential effects are calculated by using fuzzy concepts and a fuzzy reasoning system. The questions can be of two different types, e.g., qualitative and quantitative. The former is typically descriptive and it is not used by the fuzzy decision system in the reasoning process. On the contrary, the latter is used by the fuzzy engine and can be either an item chosen within a limited number of possible replies or a numeric or a boolean value. Starting from the conceptual model described in Fig. 1 (Sorlini et al., 2003), an electronic questionnaire has been developed. The questionnaire is complex and it is performed by ISPRA experts. The evaluation process is dynamic and the potential impact that can achieve one or more receptors through a set of migration paths is decided by the notifier answering the questions step by step. Therefore it is necessary to verify the correctness of the questionnaire. The evaluation processes can be described by using a directed graph, where each node represents a question and each edge points to a possible next question. The directed graph must be connected and acyclic. The for-
179
1 Institute for Environmental Protection and Research (ISPRA), is governed by the Italian Ministry of the Environment, is in charge of the evaluation of GMP risks.
Please cite this article in press as: Camastra, F., et al. A fuzzy decision system for genetically modified plant environmental risk assessment using Mamdani inference. Expert Systems with Applications (2014), http://dx.doi.org/10.1016/j.eswa.2014.09.041
148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177
180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207
ESWA 9580
No. of Pages 7, Model 5G
15 October 2014 F. Camastra et al. / Expert Systems with Applications xxx (2014) xxx–xxx
3
Fig. 1. Conceptual model. 208 209 210 211 212 213 214 215
mer requirement is necessary in order to guarantee that each question can be reachable by a path; the latter one must be fulfilled in order that the questionnaire can always end. To verify these conditions it was developed a tool that controls that the directed graph, yielded by the questionnaire, is connected and acyclic. The resulting questionnaire graph consists of 362 questions, and some parts of graph are reported in Fig. 2. The notifier’s answers to the electronic questionnaire are the input of the FDS.
216
4. The fuzzy decision system
217
The FDS has the same architecture of a Fuzzy Logic Control System. The fuzzy inference system (FIS), based on Mamdani inference (Lin & Lee, 1996; Nedjah & de Macedo Mourelle, 2006), of FDS has been implemented using the jFuzzyLogic library. Therefore the section is organized in two subsections. In the former subsection we discuss the architecture of a generic Fuzzy Logic Control System and, hence, of our system. In the latter subsection jFuzzyLogic library is surveyed.
218 219 220 221 222 223 224
225
4.1. A Fuzzy Logic Control System
forming crisp measured data (e.g., number of insert copies is 1) into suitable linguistic values, namely the data becomes, for instance, number of insert copies is low. The fuzzy rule base stores the operation knowledge of the process of the domain experts. The inference engine is the FLC core, and can simulate human decision making process by performing approximate reasoning in order to achieve a desired control strategy. The defuzzifier is used for yielding a control action inferred by the inference engine. In the inference engine the generalized modus ponens (Lin & Lee, 1996) plays an important role. For the application of fuzzy reasoning in FLCs, the generalized modus ponens can written as
227 228 229 230 231 232 233 234 235 236
A Fuzzy Logic Control (FLC) system incorporates the knowledge and experience of a human operator, the so-called expert, in the design of a system that controls a process whose input–output relationships are described by a set of fuzzy control rules, e.g., IF–THEN rules. We recall that the antecedent is the part of rule delimited by the keywords IF and THEN. Whereas the consequent is the part of the rule that follows the keyword THEN. The rules involve linguistic variables (LVs) that express qualitative high level concepts. A typical FLC architecture is composed of four principal components: a fuzzifier, a fuzzy rule base, an inference engine and a defuzzifier (Lin & Lee, 1996). The fuzzifier has the task of trans-
238 239 240 241 242 243 244 245 246 247
248
Premise 1 : IF x is A THEN y is B Premise 2 : x is A0
ð1Þ
y is B0
Conclusion
250
where A; A0 ; B and B0 are fuzzy predicates i.e., Fuzzy Sets or Fuzzy Relations. In general, a fuzzy control rule, e.g., premise 1, is a fuzzy relation expressed by a fuzzy implication R ¼ A ! B. Let max-H a composition operation, conclusion B0 in Eq. (1) can be obtained as follows 0
226
237
0
0
B ¼ A HR ¼ A HðA ! BÞ
ð2Þ
where H denotes t-norm operators (e.g., min, product, Lukasiewicz). For the fuzzy implication A ! B there are several distinct fuzzy implication functions described in literature depending on the used t-norm. For instance, Mamdani min fuzzy implication is obtained by using the intersection operator in the fuzzy conjunction and a Larsen product fuzzy implication is obtained by using an algebraic product (Nedjah & de Macedo Mourelle, 2006). Moreover, generalizations of t-norms (Ciaramella, Tagliaferri, & Pedrycz, 2005) and their application for inference systems (Ciaramella, Tagliaferri, & Pedrycz, 2004) have been studied.
Please cite this article in press as: Camastra, F., et al. A fuzzy decision system for genetically modified plant environmental risk assessment using Mamdani inference. Expert Systems with Applications (2014), http://dx.doi.org/10.1016/j.eswa.2014.09.041
251 252 253 254 255 256
257 259 260 261 262 263 264 265 266 267 268 269
ESWA 9580
No. of Pages 7, Model 5G
15 October 2014 4
F. Camastra et al. / Expert Systems with Applications xxx (2014) xxx–xxx
Fig. 2. Parts of the questionnaire graph. The whole graph is too large to be fully shown. 270 271 272 273 274 275
We are aware that alternative and more sophisticated approaches than Mamdani fuzzy inference are available, e.g., (Liu, Yang, Wang, Sii, & Wang, 2004, 2005; Yang, Liu, Wang, Sii, & Wang, 2006). Hence, Mamdani fuzzy inference has been chosen since is the simplest approach, adopting implicitly an Occam’s razor heuristic.
In most general cases, the fuzzy rule base has the form of a Multi-Input–Multi-Output (MIMO) system. In this case, the inference rules are combined by using the connectives AND and ELSE that can be interpreted as the intersection and the union for different definitions of fuzzy implications, respectively. For instance, if we consider the LV number of insert copies as shown in Fig. 3, the
Fig. 3. Membership functions of the linguistic variable number of insert copies.
Fig. 4. Membership functions of the linguistic variable potential risk of the insert.
Please cite this article in press as: Camastra, F., et al. A fuzzy decision system for genetically modified plant environmental risk assessment using Mamdani inference. Expert Systems with Applications (2014), http://dx.doi.org/10.1016/j.eswa.2014.09.041
276 277 278 279 280 281
ESWA 9580
No. of Pages 7, Model 5G
15 October 2014 5
F. Camastra et al. / Expert Systems with Applications xxx (2014) xxx–xxx 282 283
fuzzy inference system of the LVs number of insert copies and number of introduced sequences could be represented by:
284
IF number of insert copies is Low AND number of introduced sequences is High THEN potential risk of the insert is High ELSE IF number of insert copies is High AND number of introduced sequences is Medium THEN
ð3Þ
potential risk of the insert is High ELSE IF number of insert copies is Low AND number of introduced sequences is Low THEN 286
290
291
4.2. jFuzzyLogic
292
jFuzzyLogic (Cingolani et al., 2012) is an open source software library for fuzzy systems which allows designing fuzzy logic controllers supporting the standard Fuzzy Control Programming (IEC, 2000). The library is written in Java and permits FLC design and implementation, following IEC standard for Fuzzy Control Language (FCL). jFuzzyLogic allows implementing a fuzzy inference system (FIS). A FIS is usually composed of one or more Function Blocks (FBs). Each FB has variables (input, output or instances) and one or more Rule Blocks (RBs). Each RB is composed of a set of rules, as well as Aggregation, Activation and Accumulation methods. Having said that, jFuzzyLogic uses ANTLR (Parr & Quong, 1995) that generates a lexer and a parser based on a FCL grammar defined by the user. The parser uses a left to right leftmost derivation recursive strategy, formally known as LL(⁄). Using the lexer and the parser generated by ANTLR, it can parse FCL files by creating an Abstract Syntax Tree (AST). The AST is converted into an Interpreter Syntax Tree (IST), which is capable of performing the required computations. This means that the IST can represent the grammar, in a similar way of AST, but it is also capable of performing computations. Moreover, all methods defined in the IEC norm are implemented in jFuzzyLogic. It should be noted that jFuzzyLogic fulfills the definitions of Aggregation, Activation and Accumulation as defined in (IEC, 2000). Aggregation methods define the t-norms and t-conorms corresponding to the familiar AND, OR and NOT operators (Ciaramella et al., 2005). These can be Minimum, Product or Bounded difference operators. Each set of operators fulfills De Morgan’s laws (Cormen, Leiserson, Rivest, & Stein, 2009). The Activation method establishes how rule antecedent modifies the consequent. Activation operators are Minimum and Product. Accumulation methods determine how the consequents from multiple rules in the same RB are combined. Accumulation methods (IEC, 2000) include Maximum, Bounded sum, Normed sum, Probabilistic OR and Sum. Only two membership functions are defined in the IEC standard, i.e., singleton and piecewise linear. jFuzzyLogic implements other membership functions such as trapezoidal, sigmoidal, gaussian, generalized bell, difference of sigmoidal, and cosine, too. Moreover, jFuzzyLogic permits making arbitrary memberships by means of a combination of mathematical functions. Thanks to the flexibility in defining membership functions that can be sampled in a fixed number of points, called samples. The number, that is by default one thousand, can be modified on the basis of the precision-speed trade-off required for a
288 289
293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333
334
5. System validation
348
The FDS has a knowledge base organized in 123 FBs and it consists of 6215 rules of the type described in Section 3. The knowledge base has been designed by brainstorming, involving all the authors of the work. The author team was formed by agrobiologists and computer scientists. The formers with no knowledge in fuzzy reasoning, the others with no skills in risk assessment. The design and the debugging of the whole knowledge base has required several months. The knowledge base is available on request for further investigations and comparisons. FDS was tested producing 220 ERAs related to GMPs, namely Bt-maize, i.e., maize genetically modified by using Bacillus thuringiensis toxin (Sorlini et al., 2003), and oilseed rape. We tested different configurations for the inference system, e.g., different t-norms, t-conorms, generalization of norms. The results were similar and therefore, on the basis of Occam’s razor, we use the Mamdani fuzzy system. Besides, this system has the not negligible advantage to be understood by non-experts in fuzzy reasoning, how some of the authors, the agrobiologists, are. In the rest of the section we present the results obtained by a Mamdani fuzzy inference system using min and max for t-norm and t-conorm operators, respectively and a Center of area defuzzification mechanism. The ERAs, yielded by FDS, were submitted to a pool of ISPRA experts, not involved in the FDS knowledge base definition, in order to assess the consistency and completeness of FDS evaluations. For instance, we describe in Table 1 two ERAs taken from the 220 ERAs produced by FDS. The ERAs have been yielded by two antithetical scenarios. In the former all risks about GMP cultivation are negligible. Instead, in the latter some risks are potentially high. A further system validation has been performed applying the FDS at trial case studies discussed in (Lener et al., 2013). In that case studies, maize and oilseed rape have been cultivated in representative potential receiving environments for the chosen species. The presence of GMPs, e.g., herbicide tolerant oilseed rape and insect resistant MON 810 maize have been only simulated, since in Italy the GMP cultivation is not allowed by law. The study sites are located in four Italian regions, i.e., Apulia, Emilia-Romagna, Lazio and Sicily. In two cases maize and oilseed rape have been cultivated at the same location while in one site
349
335 336 337 338 339 340 341 342 343 344 345 346 347
potential risk of the insert is Low In Fig. 4 an example of fuzzy memberships for the output LV potential risk of the insert is presented. Finally, a crisp (i.e., non fuzzy) output is obtained considering a Center of Area defuzzifier (Lin & Lee, 1996).
287
specific application. Inference is performed by evaluating membership functions at these samples. To perform the sampling, the domain, called universe, of each variable, must be estimated. The variable universe is defined as the range where the variable assumes non-negligible value. For each variable, both membership function and term are considered during the universe computation. After the examination of all rules is terminated, the accumulation for each variable is complete. The last step in the evaluation of a FIS is the defuzzification (Lin & Lee, 1996). The value for each variable is computed by means of the defuzzification method selected by the user within the following set: Center of gravity, Rightmost Max, Center of area, Leftmost Max, Mean max (for the continuous membership functions), or Center of gravity (for the discrete membership functions).
Table 1 Estimation of the risks for the first and the second scenarios. Risk Potential Potential Potential Potential
expression of the inserts risk of the insert risks of the sequences risk of toxicity
First scenario
Second scenario
Medium Low Low Low
Medium Low High Low
Please cite this article in press as: Camastra, F., et al. A fuzzy decision system for genetically modified plant environmental risk assessment using Mamdani inference. Expert Systems with Applications (2014), http://dx.doi.org/10.1016/j.eswa.2014.09.041
350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387
ESWA 9580
No. of Pages 7, Model 5G
15 October 2014 6
Q4
F. Camastra et al. / Expert Systems with Applications xxx (2014) xxx–xxx
Table 2 Identified risks for oilseed rape and maize field trials. Risk of
Oilseed rape
Maize
Allergenic effects on operators Allergenic effects on human population Potential toxicity for pollen consumers Food and feed chain contamination Pollution of natural genetic resources Increasing weed population Changes to biodiversity Changes to seed consumer population Changes to soil microbe and fungus diversity Changes in crop residues detrivorous population Changes to agricultural practice Changes to propagation organ consumer population Changes to structure of symbiotic population Changes to target pathogen host range Changes to consumer populations Development of resistant target consumers Development of new pathogens Changes to structure of rhizosphere population
Low Low Low Medium Medium High Low Low High High Low Low
Low Low High Low Low Low Low Low High High Low Low
Medium Low Low Low Low High
Medium Low Medium Medium Low Medium
404
only maize has been studied, thus seven different scenarios have been obtained. The different risk scenarios have been assessed on a case-by-case basis applying our FDS. Independently from the sites where the crops have been cultivated, the same potential risks have been identified. This is justifiable on the basis of the following two main reasons. The former reason is that potential receptors depend on the crop characteristics. The latter one is that in Italy the receiving environment where these crops are cultivated are similar, when it considers the potential target. Besides, part of the similitudes can be explained by a missing knowledge of some environmental characteristics, that the FDS, adopting a precautionary approach, the so-called worst-case scenario, assumes to be as worst as possible, as recommended in the European Community Directive. The resulting identified potential risks for maize and oilseed rape, reported in Table 2, have been considered coherent and consistent by a pool of experts not involved in FDS design and development.
405
6. Conclusions
406
In this paper a FDS for the environmental risk assessment of GMPs has been presented. The risk assessment in the FDS is obtained by using Mamdani fuzzy inference. Mamdani inference was chosen since it is the simplest inference method and the experimentation of other alternative fuzzy inference techniques produced similar results. The decisions yielded by FDS have been validated on real world cases by the human experts, not involved in FDS design, confirming, in this way, the reliability and correctness of FDS decisions. Main FDS limitation consists in the time, e.g. several months, required to develop the knowledge base of the system. Whereas main advantages are the novelty of the approach to the GMP environmental risk assessment and the automation of ERA process that, at present, involves the management of several paper documents. In the next future, we plan to carry out two different research lines. Firstly, we are going to develop machine learning algorithms that allow the learning of the knowledge base of FDS, starting from the human expert reports. To this purpose, preliminary studies, obtained using genetic techniques, e.g., eXtended Classifier System (XCS) (Urbanowicz & Moore, 2009), seems to be promising. Secondly, we will investigate generalizations of fuzzy inference methodologies recently introduced. In particular, we plan to
388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403
407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428
deepen the application to FDS system of the generic rule-base inference methodology using the evidential reasoning (Liu et al., 2004, Liu, Yang, Wang, & Sii, 2005; Yang et al., 2006), that has arisen interest in the scientific community for its performance and possible applications. Moreover, we are going to study the possible application, to FDS system, of the fuzzy rule-based bayesian reasoning approach (Yang et al., 2008).
429
Acknowledgements
436
Firstly, the authors are indebted to the anonymous reviewers for their valuable comments. This research has been partially funded by the LIFE project MAN-GMP-ITA (Agreement n. LIFE08 NAT/IT/000334). The research was entirely developed when Alfredo Starace was at the Department of Science and Technology, University of Naples Parthenope.
437
References
443
Bajpai, S., Sachdeva, A., & Gupta, J. (2010). Security risk assessment: Applying the concept of fuzzy logic. Journal of Hazardous Materials, 173, 258–264. Bukhari, A., Tusseyeva, I., Lee, B.-G., & Kim, Y.-G. (2013). An intelligent real-time multi-vessel collision risk assessment system from VTS view point based on fuzzy inference system. Expert Systems with Applications, 40, 1220–1230. Camastra, F., Ciaramella, A., & Staiano, A. (2014). A note on some mathematical models on the effects of Bt-maize exposure. Environmental and Ecological Statistics, 21, 477–485. Chen, Y.-L., & Weng, C.-H. (2009). Mining fuzzy association rules from questionnaire data. Knowledge-Based Systems, 22, 46–56. Chen, Z., Zhao, L., & Lee, K. (2010). Environmental risk assessment of offshore produced water discharges using a hybrid fuzzy-stochastic modeling approach. Environmental Modelling & Software, 25, 782–792. Cho, H.-N., Choi, H.-H., & Kim, Y.-B. (2002). A risk assessment methodology for incorporating uncertainties using fuzzy concepts. Reliability Engineering and System Safety, 78, 173–183. Chongfu, H. (1996). Fuzzy risk assessment of urban natural hazards. Fuzzy Sets and Systems, 83, 271–282. Ciaramella, A., Tagliaferri, R., & Pedrycz, W. (2004). Inference systems by using ordinal sums and genetic algorithms. In IEEE annual meeting of the fuzzy information, 2004. Processing NAFIPS ’04 (Vol. 2, pp. 629–634). IEEE. Ciaramella, A., Tagliaferri, R., & Pedrycz, W. (2005). The genetic development of ordinal sums. Fuzzy Sets and Systems, 151, 303–325. Cingolani & Alcala-Fdez, J. (2012). jFuzzyLogic: A robust and flexible Fuzzy-logic inference system language implementation. In Proceedings of IEEE world congress on computational intelligence 2012 (pp. 1–8). IEEE. Cormen, T., Leiserson, C., Rivest, R., & Stein, C. (2009). Introduction to algorithms. The MIT Press. Darbra, R., Demichela, M., & Mure, S. (2008). Preliminary risk assessment of ecotoxic substances accidental releases in major risk installations through fuzzy logic. Process Safety and Environment Protection, 86, 103–111. Davidson, V., Ryks, J., & Fazil, A. (2006). Fuzzy risk assessment tool for microbial hazards in food systems. Fuzzy Sets and Systems, 157, 1201–1210. Dikmen, I., Birgonul, M., & Han, S. (2007). Using fuzzy risk assessment to rate cost overrun risk in international construction projects. International Journal of Project Management, 25, 494–505. Elsayed, T. (2009). Fuzzy inference system for the risk assessment of liquefied natural gas carriers during loading/offloading at terminals. Applied Ocean Research, 31, 179–185. Guimara, A., & Lapa, C. (2007). Fuzzy inference to risk assessment on nuclear engineering systems. Applied Soft Computing, 7, 17–28. Hadjimichael, M. (2009). A fuzzy expert system for aviation risk assessment. Expert System with Application, 36, 6512–6519. Harris, T. (2013). Quantitative credit risk assessment using support vector machine: Broad versus narrow default definitions. Expert Systems with Applications, 40, 4404–4413. Huang, G., Chen, Z., Tontiwachwuthikul, P., & Chakma, A. (2010). Environmental risk assessment for underground storage tanks through an interval parameter fuzzy relation analysis approach. Energy Sources, 21, 75–96. IEC (2000). International Electrotechnical Commission technical committee industrial process measurement and control. IEC 61131 – Programmable controllers. Part 7: fuzzy control programming., Technical report, International electrotechnical commission. Kahraman, C., & Kaya, I. (2009). Fuzzy process accuracy index to evaluate risk assessment of drought effects in turkey. Human and Ecological Risk Assessment, 15, 789–810. Karimi, I., & Hullermeier, E. (2007). Risk assessment system of natural hazards: A new approach based on fuzzy probability. Fuzzy Sets and Systems, 158, 987–999. Lang, A., Brunzel, S., Dolek, M., Otto, M., & Thei, B. (2011). Modelling in the light of key parameters: A call to exercise caution in field prediction of Bt-maize effects. Proceedings of the Royal Society B, Biological Sciences, 278, 980–981.
444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504
Please cite this article in press as: Camastra, F., et al. A fuzzy decision system for genetically modified plant environmental risk assessment using Mamdani inference. Expert Systems with Applications (2014), http://dx.doi.org/10.1016/j.eswa.2014.09.041
430 431 432 433 434 435
438 439 440 441 442
ESWA 9580
No. of Pages 7, Model 5G
15 October 2014 F. Camastra et al. / Expert Systems with Applications xxx (2014) xxx–xxx 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540
Lener, M., Giovannelli, V., Arpaia, S., Burgio, G., Canfora, L., Dinelli, G., et al. (2013). Applying an operating model for the environmental risk assessment in italian sites of community importance (sci) of the european commission habitats directive (92/43/EEC). Bulletin of Insectology, 66, 257–267. Li, J., Huang, G., Maqsood, I., & Huang, Y. (2007). An integrated fuzzy-stochastic modeling approach for risk assessment of groundwater contamination. Journal of Environmental Management, 82, 173–188. Lin, C.-T., & Lee, C. (1996). Neural fuzzy systems: A neuro-fuzzy synergism to intelligent systems. Prentice Hall. Liu, J., Yang, J.-B., Wang, J., & Sii, H.-S. (2005). Engineering system safety analysis and synthesis using the fuzzy rule-based evidential reasoning approach. Quality and Reliability Engineering International, 21, 387–411. Liu, J., Yang, J.-B., Wang, J., Sii, H.-S., & Wang, Y.-M. (2004). Fuzzy rule-based evidential reasoning approach for safety analysis. International Journal of General Systems, 33, 183–204. Li, W., Zhou, J., Xie, K., & Xiong, X. (2008). Power system risk assessment using a hybrid method of fuzzy set and monte carlo simulation. IEEE Transactions on Power Systems, 23. Mokhtari, K., Ren, J., Roberts, C., & Wang, J. (2012). Decision support framework for risk management on sea ports and terminals using fuzzy set theory and evidential reasoning approach. Expert Systems with Applications, 39, 5087–5103. Nedjah, N., & de Macedo Mourelle, L. (2006). Fuzzy systems engineering. Springer. Ngai, E., & Wat, F. (2003). Design and development of a fuzzy expert system for hotel selection. Omega, 31, 275–286. Oreski, S., & Oreski, G. (2014). Genetic algorithm-based heuristic for feature selection in credit risk assessment. Expert Systems with Applications, 41, 951–961. Parr, T., & Quong, R. (1995). ANTLR: A predicated-ll(k) parser generator. Software: Practice and Experience, 25, 789–810. Perry, J., Devos, Y., Arpaia, S., Bartsch, D., Gathmann, A., Hails, R., et al. (2010). A mathematical model of exposure of non-target lepidoptera to Bt-maize pollen expressing Cry1Ab within Europe. Proceedings of the Royal Society B, Biological Sciences, 277, 1417–1425. Petrovic, D., Tanasijevic, M., Milic, V., Lilic, N., Stojadinovic, S., & Svrkota, S. (2014). Risk assessment model of mining equipment failure based on fuzzy logic. Expert Systems with Applications, 41, 8157–8164.
7
Sadiq, R., & Husain, T. (2005). A fuzzy-based methodology for an aggregative environmental risk assessment: A case study of drilling waste. Environmental Modelling & Software, 20, 33–46. Samantra, C., Datta, S., & Shankar Manapantra, S. (2014). Risk assessment in it outsourcing using fuzzy decision-making approach: An indian perspective. Expert Systems with Applications, 41, 4010–4022. Smith, E., & Eloff, J. (2000). Cognitive fuzzy modeling for enhanced risk assessment in a health care institution. IEEE Intelligent Systems and Their Applications, 15, 69–75. Sorlini, C., Buiatti, M., Burgio, G., Cellini, F., Giovannelli, V., & Lener, M., et al. (2013). La valutazione del rischio ambientale dell’ immissione deliberata nell’ ambiente di organismi geneticamente modificati (in Italian), Technical report, Italian ministry of environment, http://bch.minambiente.it/EN/Biosafety/propmet.asp, 2003. Urbanowicz, R., & Moore, J. (2009). Learning classifier systems: A complete introduction, review, and roadmap. Journal of Artificial Evolution and Applications, 2009, 125. Wang, F., Ding, L., Luo, H., & Love, P. (2014). Probabilistic risk assessment of tunneling-induced damage to existing properties. Expert Systems with Applications, 41, 951–961. Wang, Y.-M., & Elhag, T. (2008). An adaptive neuro-fuzzy inference system for bridge risk assessment. Expert Systems with Applications, 34, 3099–3106. Yang, Z., Bonsall, S., & Wang, J. (2008). Fuzzy rule-based bayesian reasoning approach for prioritization of failures in FMEA. IEEE Transactions on Reliability, 57, 517–528. Yang, J.-B., Liu, J., Wang, J., Sii, H.-S., & Wang, H.-W. (2006). Belief rule-base inference methodology using the evidential reasoning approach – RIMER. IEEE Transactions on Systems, Man, and Cybernetics – Part A: Systems and Humans, 26, 266–285. Zeng, J., An, M., & Smith, N. (2007). Application of a fuzzy based decision making methodology to construction project risk assessment. International Journal of Project Management, 25, 589–600. Zhou, J. (2010). Spa-fuzzy method based real-time risk assessment for major hazard installations storing flammable gas. Safety Science, 48, 819–822.
Please cite this article in press as: Camastra, F., et al. A fuzzy decision system for genetically modified plant environmental risk assessment using Mamdani inference. Expert Systems with Applications (2014), http://dx.doi.org/10.1016/j.eswa.2014.09.041
541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575
No. of Pages 7, Model 5G
15 October 2014 Expert Systems with Applications xxx (2014) xxx–xxx 1
Contents lists available at ScienceDirect
Expert Systems with Applications journal homepage: www.elsevier.com/locate/eswa 5 6
A fuzzy decision system for genetically modified plant environmental risk assessment using Mamdani inference
3 4 7
Q1
8
Francesco Camastra a,⇑, Angelo Ciaramella a, Valeria Giovannelli b, Matteo Lener b, Valentina Rastelli b, Antonino Staiano a, Giovanni Staiano b, Alfredo Starace c a
9 10 11
b c
12 13
Department of Science and Technology, University of Naples Parthenope, Centro Direzionale Isola C4, 80143 Naples, Italy Nature Protection Department, Institute for Environmental Protection and Research (ISPRA), via v. Brancati 48, 00144 Roma, Italy Oracle Ireland Ltd., Block C, East Point Business Park, Dublin 3, Ireland
a r t i c l e
1 2 5 5 16 17
i n f o
Article history: Available online xxxx
18 19 20 21 22 23 24
Keywords: Fuzzy decision system Mamdani fuzzy inference Risk assessment Genetically modified plants Fuzzy Control Language
a b s t r a c t Environmental risk assessment (ERA) of the deliberate release of genetically modified plants (GMPs) is currently performed by human experts on the basis of own personal experience and knowledge. In this paper we describe a fuzzy decision system (FDS) for the ERA of GMPs, based on Mamdani fuzzy inference. The risk assessment in the FDS is obtained by using a fuzzy inference system (FIS), performed using jFuzzyLogic library. The FDS permits obtaining an evaluation process for the identification of potential impacts that can achieve one or more receptors through a set of migration paths. The decisions derived by FDS have been validated on real world cases by the human experts that are in charge of ERA. They have confirmed the reliability and correctness of the fuzzy system decisions. Ó 2014 Elsevier Ltd. All rights reserved.
26 27 28 29 30 31 32 33 34 35 36
37 38
1. Introduction
39
The development of genetic engineering in the last years produced a very high number of genetically modified plants (GMPs). Whereas in USA the use of GMPs is widely spread in agriculture, in Europe there are discordant policies w.r.t. GMPs. For instance, commercialization of food and feed containing or consisting of GMPs is duly approved in European Community (EC), while cultivation of new genetically modified crops are not adopted. The maize MON 810, approved by the old EC legislation framework, is currently the unique GMP cultivated in the EC (e.g., Czech Republic, Poland, Spain, Portugal, Romania and Slovakia), although the evaluations about its effect on the surrounding environment are discordant (Camastra, Ciaramella, & Staiano, 2014; Lang, Brunzel, Dolek, Otto, & Thei, 2011; Perry et al., 2010). According to EC, the environmental release of GMPs is ruled by Directive 200118EC and Regulation 18292003EC. The Directive refers to the deliberate release into the environment of GMPs and sets out two regulatory regimes: Part C for the placing on the market and Part B for the deliberate release for any other purpose, i.e., field
40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56
Q2
⇑ Corresponding author. E-mail addresses: [email protected] (F. Camastra), angelo.ciaramella@uni parthenope.it (A. Ciaramella), [email protected] (V. Giovannelli), [email protected] (M. Lener), [email protected] (V. Rastelli), [email protected] (A. Staiano), giovanni.staiano@ isprambiente.it (G. Staiano), [email protected] (A. Starace).
trials (Sorlini et al., 2003). In both legislations the notifier, i.e., the person who requests the release into the environment of GMP, must perform an environmental risk assessment (ERA) on the issue. The ERA is formally defined as ‘‘the evaluation of risks to human health and the environment, whether direct or indirect, immediate or delayed, which the deliberate release or the placing on the market of GMPs may pose’’. ERA should be carried out case by case, meaning that its conclusion may depends on the GMPs and trait concerned, their intended uses, and the potential receiving environments. The ERA process should lead to the identification and evaluation of potential adverse effects of the GMP, and, at the same time, it should be conducted with a view for identifying if there is a need for risk management and it should provide the basis for the monitoring plans. The aim of this work is the development of a fuzzy decision system that should advise the notifier in performing the ERA about the cultivation of a specific GMP. ERA process is often performed in presence of incomplete and imprecise data. Moreover, it is generally yielded using the personal experience and knowledge of the notifier. Therefore the usage of fuzzy reasoning in the ERA is particularly appropriate as witnessed by the extensive application of fuzzy reasoning to the risk assessment in disparate fields (Bukhari, Tusseyeva, Lee, & Kim, 2013; Chen & Weng, 2009; Chen, Zhao, & Lee, 2010; Cho, Choi, & Kim, 2002; Davidson, Ryks, & Fazil, 2006; Guimara & Lapa, 2007; Kahraman & Kaya, 2009; Karimi & Hullermeier, 2007; Li, Zhou, Xie, & Xiong, 2008;
http://dx.doi.org/10.1016/j.eswa.2014.09.041 0957-4174/Ó 2014 Elsevier Ltd. All rights reserved.
Please cite this article in press as: Camastra, F., et al. A fuzzy decision system for genetically modified plant environmental risk assessment using Mamdani inference. Expert Systems with Applications (2014), http://dx.doi.org/10.1016/j.eswa.2014.09.041
57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82
ESWA 9580
No. of Pages 7, Model 5G
15 October 2014 2 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99
F. Camastra et al. / Expert Systems with Applications xxx (2014) xxx–xxx
Mokhtari, Ren, Roberts, & Wang, 2012; Ngai & Wat, 2003; Petrovic et al., 2014; Sadiq & Husain, 2005; Samantra, Datta, & Shankar Manapantra, 2014; Smith & Eloff, 2000; Wang & Elhag, 2008; Yang, Bonsall, & Wang, 2008). To our best knowledge, no Fuzzybased Systems for ERA of GMPs have been developed, yet. For this reason, the fuzzy decision system (FDS), object of the paper, represents a novelty. The FDS is inspired by Sorlini et al.’s methodological proposal of performing ERA on GMP field trials (Sorlini et al., 2003). The methodology allows describing the relationships between potential receptors and the harmful characteristics of a GMP field trial, leading to the identification of potential impacts. The paper is organized as follows: In Section 2 a brief overview of works, that apply fuzzy reasoning to the risk assessment, is provided; In Section 3 the methodological proposal that inspires the system is presented; The FDS structure is discussed in Section 4; Section 5 describes how the system validation has been performed; Finally some conclusions are drawn in Section 6.
100
2. Fuzzy based environmental risk assessment
101
In these last years several works have been conducted for the risk assessment in various scientific areas. Several applications were developed for the monitoring of industrial processes and their environmental impacts. Darbra et al. introduced a fuzzy logic model to assess preliminarily the risk of accidental releases of ecotoxic substances in hazard plants (Darbra, Demichela, & Mure, 2008). The tool has been tested with a set of storage yards of ecotoxic substances, mainly oil, in the Piedmont Region areas (Italy). Huang et al. proposed an interval parameter fuzzy relation analysis model for environmental risk assessment of petroleumcontaminated aquifers due to leakage from underground storage tanks (Huang, Chen, Tontiwachwuthikul, & Chakma, 2010). The modeling results have provided bases for determining desirable site remediation actions. Sadiq and Husain developed a fuzzy based methodology for estimating aggregative risk of various environmental activities, pollution sources and routes in a given process (Sadiq & Husain, 2005). The developed methodology was applied to a case study of offshore drilling waste for evaluating various discharge scenarios. Chen et al. and Li et al. applied fuzzy and statistical mechanisms for environmental risk assessment (Li, Huang, Maqsood, & Huang, 2007), e.g., pollutant dispersion for the prediction of the environmental risks (Chen et al., 2010), petroleum-contaminated groundwater system in western Canada (Li et al., 2008). Zhou proposed a set-pair analysis based on a fuzzy assessment method for the real-time monitoring and control of major hazard installations storing flammable gas (Zhou, 2010). Bajpai adopted a security risk assessment technique for chemical process industries handling hazardous chemicals in bulk (Bajpai, Sachdeva, & Gupta, 2010). Guimara and Lapa presented a nuclear case study (Guimara & Lapa, 2007), in which a fuzzy inference system was used as alternative approach in risk analysis. A standard four-loop pressurized water reactor containment cooling system was used as example case. Elsayed developed a multiple attribute risk assessment approach using a fuzzy inference system (Elsayed, 2009) and used it as an alternative approach to qualitative risk matrix techniques. The system is currently used in many industries and ship classification societies. Zeng et al. and Dikmen et al. proposed a fuzzy risk assessment methodology for construction projects (Dikmen, Birgonul, & Han, 2007; Zeng, An, & Smith, 2007). Wang and Elhag adopted a neuro-fuzzy system for bridge risks (Wang & Elhag, 2008) evaluated periodically so that the bridges with high risks can be maintained timely. Hadjimichael developed a fuzzy based tool, valuable to airline safety departments, for examining risk trends, to pilots and dispatchers for assessing risks associated with each flight, and to airline management for quantifying the effects of making safety-related changes (Hadjimichael,
102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146
2009). Kahraman and Kaya adopted a fuzzy based system as a systematic procedure for predicting potential risks to human health or the environment (Kahraman & Kaya, 2009). The proposed methodology is applied to measure drought effects by analyzing the fullness rates of the dams in Istanbul. Davidson et al. developed a fuzzy risk assessment tool for early-stage risk assessment of microbial hazards in food systems (Davidson et al., 2006). Karimi and Hullermeier, and Chongfu presented systems for assessing the risk of natural disasters (Chongfu, 1996; Karimi & Hullermeier, 2007), particularly under highly uncertain conditions. The systems has been applied for assessing the earthquake risk in metropolitan area. Recently, fuzzy based approaches are used in risk assessment model of mining equipment failure (Petrovic et al., 2014), in a real-time multi-vessel collision risk assessment system (Bukhari et al., 2013) and in risk assessment in IT outsourcing (Samantra et al., 2014). We have also to mention that approaches for risk assessment, alternative to fuzzy reasoning, have been proposed. To this purpose, we have to remark that SVM (Harris, 2013) and hybrid neural networks for credit risk assessment (Oreski & Oreski, 2014), and the Bayesian networks for risk assessment of tunneling-induced damage (Wang, Ding, Luo, & Love, 2014) have been used. Finally, we conclude the overview quoting two works, even if not strictly related to ERA, that have influenced the development of the FDS, object of the paper. The former introduced the questionnaire data mining problem and defined the rule patterns that can be mined from questionnaire data (Chen & Weng, 2009). The latter described the development of a fuzzy expert system for hotel selection (Ngai & Wat, 2003). The viability of the system as an effective procedure for hotel selection was ascertained by the positive feedback obtained from the survey questionnaires.
147
3. The methodological approach
178
The methodological proposal, that has inspired the system is based on a conceptual model developed by ISPRA1 experts (Sorlini et al., 2003). The schema, shown in Fig. 1, illustrates the possible paths of the impact from a specific source to a given receptor through disparate diffusion factors and migration routes. The model implies that the notifier compiles an electronic questionnaire. The notifier answers are collected in a relational database management system and, in a second time, become input of a fuzzy decision engine that is the system core and provides to the notifier the overall evaluation of risk assessment related to a specific GMP. The questionnaire can be grouped in specific sets of questions where each set corresponds to a specific box of the diagram of the conceptual model. For each block the potential effects are calculated by using fuzzy concepts and a fuzzy reasoning system. The questions can be of two different types, e.g., qualitative and quantitative. The former is typically descriptive and it is not used by the fuzzy decision system in the reasoning process. On the contrary, the latter is used by the fuzzy engine and can be either an item chosen within a limited number of possible replies or a numeric or a boolean value. Starting from the conceptual model described in Fig. 1 (Sorlini et al., 2003), an electronic questionnaire has been developed. The questionnaire is complex and it is performed by ISPRA experts. The evaluation process is dynamic and the potential impact that can achieve one or more receptors through a set of migration paths is decided by the notifier answering the questions step by step. Therefore it is necessary to verify the correctness of the questionnaire. The evaluation processes can be described by using a directed graph, where each node represents a question and each edge points to a possible next question. The directed graph must be connected and acyclic. The for-
179
1 Institute for Environmental Protection and Research (ISPRA), is governed by the Italian Ministry of the Environment, is in charge of the evaluation of GMP risks.
Please cite this article in press as: Camastra, F., et al. A fuzzy decision system for genetically modified plant environmental risk assessment using Mamdani inference. Expert Systems with Applications (2014), http://dx.doi.org/10.1016/j.eswa.2014.09.041
148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177
180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207
ESWA 9580
No. of Pages 7, Model 5G
15 October 2014 F. Camastra et al. / Expert Systems with Applications xxx (2014) xxx–xxx
3
Fig. 1. Conceptual model. 208 209 210 211 212 213 214 215
mer requirement is necessary in order to guarantee that each question can be reachable by a path; the latter one must be fulfilled in order that the questionnaire can always end. To verify these conditions it was developed a tool that controls that the directed graph, yielded by the questionnaire, is connected and acyclic. The resulting questionnaire graph consists of 362 questions, and some parts of graph are reported in Fig. 2. The notifier’s answers to the electronic questionnaire are the input of the FDS.
216
4. The fuzzy decision system
217
The FDS has the same architecture of a Fuzzy Logic Control System. The fuzzy inference system (FIS), based on Mamdani inference (Lin & Lee, 1996; Nedjah & de Macedo Mourelle, 2006), of FDS has been implemented using the jFuzzyLogic library. Therefore the section is organized in two subsections. In the former subsection we discuss the architecture of a generic Fuzzy Logic Control System and, hence, of our system. In the latter subsection jFuzzyLogic library is surveyed.
218 219 220 221 222 223 224
225
4.1. A Fuzzy Logic Control System
forming crisp measured data (e.g., number of insert copies is 1) into suitable linguistic values, namely the data becomes, for instance, number of insert copies is low. The fuzzy rule base stores the operation knowledge of the process of the domain experts. The inference engine is the FLC core, and can simulate human decision making process by performing approximate reasoning in order to achieve a desired control strategy. The defuzzifier is used for yielding a control action inferred by the inference engine. In the inference engine the generalized modus ponens (Lin & Lee, 1996) plays an important role. For the application of fuzzy reasoning in FLCs, the generalized modus ponens can written as
227 228 229 230 231 232 233 234 235 236
A Fuzzy Logic Control (FLC) system incorporates the knowledge and experience of a human operator, the so-called expert, in the design of a system that controls a process whose input–output relationships are described by a set of fuzzy control rules, e.g., IF–THEN rules. We recall that the antecedent is the part of rule delimited by the keywords IF and THEN. Whereas the consequent is the part of the rule that follows the keyword THEN. The rules involve linguistic variables (LVs) that express qualitative high level concepts. A typical FLC architecture is composed of four principal components: a fuzzifier, a fuzzy rule base, an inference engine and a defuzzifier (Lin & Lee, 1996). The fuzzifier has the task of trans-
238 239 240 241 242 243 244 245 246 247
248
Premise 1 : IF x is A THEN y is B Premise 2 : x is A0
ð1Þ
y is B0
Conclusion
250
where A; A0 ; B and B0 are fuzzy predicates i.e., Fuzzy Sets or Fuzzy Relations. In general, a fuzzy control rule, e.g., premise 1, is a fuzzy relation expressed by a fuzzy implication R ¼ A ! B. Let max-H a composition operation, conclusion B0 in Eq. (1) can be obtained as follows 0
226
237
0
0
B ¼ A HR ¼ A HðA ! BÞ
ð2Þ
where H denotes t-norm operators (e.g., min, product, Lukasiewicz). For the fuzzy implication A ! B there are several distinct fuzzy implication functions described in literature depending on the used t-norm. For instance, Mamdani min fuzzy implication is obtained by using the intersection operator in the fuzzy conjunction and a Larsen product fuzzy implication is obtained by using an algebraic product (Nedjah & de Macedo Mourelle, 2006). Moreover, generalizations of t-norms (Ciaramella, Tagliaferri, & Pedrycz, 2005) and their application for inference systems (Ciaramella, Tagliaferri, & Pedrycz, 2004) have been studied.
Please cite this article in press as: Camastra, F., et al. A fuzzy decision system for genetically modified plant environmental risk assessment using Mamdani inference. Expert Systems with Applications (2014), http://dx.doi.org/10.1016/j.eswa.2014.09.041
251 252 253 254 255 256
257 259 260 261 262 263 264 265 266 267 268 269
ESWA 9580
No. of Pages 7, Model 5G
15 October 2014 4
F. Camastra et al. / Expert Systems with Applications xxx (2014) xxx–xxx
Fig. 2. Parts of the questionnaire graph. The whole graph is too large to be fully shown. 270 271 272 273 274 275
We are aware that alternative and more sophisticated approaches than Mamdani fuzzy inference are available, e.g., (Liu, Yang, Wang, Sii, & Wang, 2004, 2005; Yang, Liu, Wang, Sii, & Wang, 2006). Hence, Mamdani fuzzy inference has been chosen since is the simplest approach, adopting implicitly an Occam’s razor heuristic.
In most general cases, the fuzzy rule base has the form of a Multi-Input–Multi-Output (MIMO) system. In this case, the inference rules are combined by using the connectives AND and ELSE that can be interpreted as the intersection and the union for different definitions of fuzzy implications, respectively. For instance, if we consider the LV number of insert copies as shown in Fig. 3, the
Fig. 3. Membership functions of the linguistic variable number of insert copies.
Fig. 4. Membership functions of the linguistic variable potential risk of the insert.
Please cite this article in press as: Camastra, F., et al. A fuzzy decision system for genetically modified plant environmental risk assessment using Mamdani inference. Expert Systems with Applications (2014), http://dx.doi.org/10.1016/j.eswa.2014.09.041
276 277 278 279 280 281
ESWA 9580
No. of Pages 7, Model 5G
15 October 2014 5
F. Camastra et al. / Expert Systems with Applications xxx (2014) xxx–xxx 282 283
fuzzy inference system of the LVs number of insert copies and number of introduced sequences could be represented by:
284
IF number of insert copies is Low AND number of introduced sequences is High THEN potential risk of the insert is High ELSE IF number of insert copies is High AND number of introduced sequences is Medium THEN
ð3Þ
potential risk of the insert is High ELSE IF number of insert copies is Low AND number of introduced sequences is Low THEN 286
290
291
4.2. jFuzzyLogic
292
jFuzzyLogic (Cingolani et al., 2012) is an open source software library for fuzzy systems which allows designing fuzzy logic controllers supporting the standard Fuzzy Control Programming (IEC, 2000). The library is written in Java and permits FLC design and implementation, following IEC standard for Fuzzy Control Language (FCL). jFuzzyLogic allows implementing a fuzzy inference system (FIS). A FIS is usually composed of one or more Function Blocks (FBs). Each FB has variables (input, output or instances) and one or more Rule Blocks (RBs). Each RB is composed of a set of rules, as well as Aggregation, Activation and Accumulation methods. Having said that, jFuzzyLogic uses ANTLR (Parr & Quong, 1995) that generates a lexer and a parser based on a FCL grammar defined by the user. The parser uses a left to right leftmost derivation recursive strategy, formally known as LL(⁄). Using the lexer and the parser generated by ANTLR, it can parse FCL files by creating an Abstract Syntax Tree (AST). The AST is converted into an Interpreter Syntax Tree (IST), which is capable of performing the required computations. This means that the IST can represent the grammar, in a similar way of AST, but it is also capable of performing computations. Moreover, all methods defined in the IEC norm are implemented in jFuzzyLogic. It should be noted that jFuzzyLogic fulfills the definitions of Aggregation, Activation and Accumulation as defined in (IEC, 2000). Aggregation methods define the t-norms and t-conorms corresponding to the familiar AND, OR and NOT operators (Ciaramella et al., 2005). These can be Minimum, Product or Bounded difference operators. Each set of operators fulfills De Morgan’s laws (Cormen, Leiserson, Rivest, & Stein, 2009). The Activation method establishes how rule antecedent modifies the consequent. Activation operators are Minimum and Product. Accumulation methods determine how the consequents from multiple rules in the same RB are combined. Accumulation methods (IEC, 2000) include Maximum, Bounded sum, Normed sum, Probabilistic OR and Sum. Only two membership functions are defined in the IEC standard, i.e., singleton and piecewise linear. jFuzzyLogic implements other membership functions such as trapezoidal, sigmoidal, gaussian, generalized bell, difference of sigmoidal, and cosine, too. Moreover, jFuzzyLogic permits making arbitrary memberships by means of a combination of mathematical functions. Thanks to the flexibility in defining membership functions that can be sampled in a fixed number of points, called samples. The number, that is by default one thousand, can be modified on the basis of the precision-speed trade-off required for a
288 289
293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333
334
5. System validation
348
The FDS has a knowledge base organized in 123 FBs and it consists of 6215 rules of the type described in Section 3. The knowledge base has been designed by brainstorming, involving all the authors of the work. The author team was formed by agrobiologists and computer scientists. The formers with no knowledge in fuzzy reasoning, the others with no skills in risk assessment. The design and the debugging of the whole knowledge base has required several months. The knowledge base is available on request for further investigations and comparisons. FDS was tested producing 220 ERAs related to GMPs, namely Bt-maize, i.e., maize genetically modified by using Bacillus thuringiensis toxin (Sorlini et al., 2003), and oilseed rape. We tested different configurations for the inference system, e.g., different t-norms, t-conorms, generalization of norms. The results were similar and therefore, on the basis of Occam’s razor, we use the Mamdani fuzzy system. Besides, this system has the not negligible advantage to be understood by non-experts in fuzzy reasoning, how some of the authors, the agrobiologists, are. In the rest of the section we present the results obtained by a Mamdani fuzzy inference system using min and max for t-norm and t-conorm operators, respectively and a Center of area defuzzification mechanism. The ERAs, yielded by FDS, were submitted to a pool of ISPRA experts, not involved in the FDS knowledge base definition, in order to assess the consistency and completeness of FDS evaluations. For instance, we describe in Table 1 two ERAs taken from the 220 ERAs produced by FDS. The ERAs have been yielded by two antithetical scenarios. In the former all risks about GMP cultivation are negligible. Instead, in the latter some risks are potentially high. A further system validation has been performed applying the FDS at trial case studies discussed in (Lener et al., 2013). In that case studies, maize and oilseed rape have been cultivated in representative potential receiving environments for the chosen species. The presence of GMPs, e.g., herbicide tolerant oilseed rape and insect resistant MON 810 maize have been only simulated, since in Italy the GMP cultivation is not allowed by law. The study sites are located in four Italian regions, i.e., Apulia, Emilia-Romagna, Lazio and Sicily. In two cases maize and oilseed rape have been cultivated at the same location while in one site
349
335 336 337 338 339 340 341 342 343 344 345 346 347
potential risk of the insert is Low In Fig. 4 an example of fuzzy memberships for the output LV potential risk of the insert is presented. Finally, a crisp (i.e., non fuzzy) output is obtained considering a Center of Area defuzzifier (Lin & Lee, 1996).
287
specific application. Inference is performed by evaluating membership functions at these samples. To perform the sampling, the domain, called universe, of each variable, must be estimated. The variable universe is defined as the range where the variable assumes non-negligible value. For each variable, both membership function and term are considered during the universe computation. After the examination of all rules is terminated, the accumulation for each variable is complete. The last step in the evaluation of a FIS is the defuzzification (Lin & Lee, 1996). The value for each variable is computed by means of the defuzzification method selected by the user within the following set: Center of gravity, Rightmost Max, Center of area, Leftmost Max, Mean max (for the continuous membership functions), or Center of gravity (for the discrete membership functions).
Table 1 Estimation of the risks for the first and the second scenarios. Risk Potential Potential Potential Potential
expression of the inserts risk of the insert risks of the sequences risk of toxicity
First scenario
Second scenario
Medium Low Low Low
Medium Low High Low
Please cite this article in press as: Camastra, F., et al. A fuzzy decision system for genetically modified plant environmental risk assessment using Mamdani inference. Expert Systems with Applications (2014), http://dx.doi.org/10.1016/j.eswa.2014.09.041
350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387
ESWA 9580
No. of Pages 7, Model 5G
15 October 2014 6
Q4
F. Camastra et al. / Expert Systems with Applications xxx (2014) xxx–xxx
Table 2 Identified risks for oilseed rape and maize field trials. Risk of
Oilseed rape
Maize
Allergenic effects on operators Allergenic effects on human population Potential toxicity for pollen consumers Food and feed chain contamination Pollution of natural genetic resources Increasing weed population Changes to biodiversity Changes to seed consumer population Changes to soil microbe and fungus diversity Changes in crop residues detrivorous population Changes to agricultural practice Changes to propagation organ consumer population Changes to structure of symbiotic population Changes to target pathogen host range Changes to consumer populations Development of resistant target consumers Development of new pathogens Changes to structure of rhizosphere population
Low Low Low Medium Medium High Low Low High High Low Low
Low Low High Low Low Low Low Low High High Low Low
Medium Low Low Low Low High
Medium Low Medium Medium Low Medium
404
only maize has been studied, thus seven different scenarios have been obtained. The different risk scenarios have been assessed on a case-by-case basis applying our FDS. Independently from the sites where the crops have been cultivated, the same potential risks have been identified. This is justifiable on the basis of the following two main reasons. The former reason is that potential receptors depend on the crop characteristics. The latter one is that in Italy the receiving environment where these crops are cultivated are similar, when it considers the potential target. Besides, part of the similitudes can be explained by a missing knowledge of some environmental characteristics, that the FDS, adopting a precautionary approach, the so-called worst-case scenario, assumes to be as worst as possible, as recommended in the European Community Directive. The resulting identified potential risks for maize and oilseed rape, reported in Table 2, have been considered coherent and consistent by a pool of experts not involved in FDS design and development.
405
6. Conclusions
406
In this paper a FDS for the environmental risk assessment of GMPs has been presented. The risk assessment in the FDS is obtained by using Mamdani fuzzy inference. Mamdani inference was chosen since it is the simplest inference method and the experimentation of other alternative fuzzy inference techniques produced similar results. The decisions yielded by FDS have been validated on real world cases by the human experts, not involved in FDS design, confirming, in this way, the reliability and correctness of FDS decisions. Main FDS limitation consists in the time, e.g. several months, required to develop the knowledge base of the system. Whereas main advantages are the novelty of the approach to the GMP environmental risk assessment and the automation of ERA process that, at present, involves the management of several paper documents. In the next future, we plan to carry out two different research lines. Firstly, we are going to develop machine learning algorithms that allow the learning of the knowledge base of FDS, starting from the human expert reports. To this purpose, preliminary studies, obtained using genetic techniques, e.g., eXtended Classifier System (XCS) (Urbanowicz & Moore, 2009), seems to be promising. Secondly, we will investigate generalizations of fuzzy inference methodologies recently introduced. In particular, we plan to
388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403
407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428
deepen the application to FDS system of the generic rule-base inference methodology using the evidential reasoning (Liu et al., 2004, Liu, Yang, Wang, & Sii, 2005; Yang et al., 2006), that has arisen interest in the scientific community for its performance and possible applications. Moreover, we are going to study the possible application, to FDS system, of the fuzzy rule-based bayesian reasoning approach (Yang et al., 2008).
429
Acknowledgements
436
Firstly, the authors are indebted to the anonymous reviewers for their valuable comments. This research has been partially funded by the LIFE project MAN-GMP-ITA (Agreement n. LIFE08 NAT/IT/000334). The research was entirely developed when Alfredo Starace was at the Department of Science and Technology, University of Naples Parthenope.
437
References
443
Bajpai, S., Sachdeva, A., & Gupta, J. (2010). Security risk assessment: Applying the concept of fuzzy logic. Journal of Hazardous Materials, 173, 258–264. Bukhari, A., Tusseyeva, I., Lee, B.-G., & Kim, Y.-G. (2013). An intelligent real-time multi-vessel collision risk assessment system from VTS view point based on fuzzy inference system. Expert Systems with Applications, 40, 1220–1230. Camastra, F., Ciaramella, A., & Staiano, A. (2014). A note on some mathematical models on the effects of Bt-maize exposure. Environmental and Ecological Statistics, 21, 477–485. Chen, Y.-L., & Weng, C.-H. (2009). Mining fuzzy association rules from questionnaire data. Knowledge-Based Systems, 22, 46–56. Chen, Z., Zhao, L., & Lee, K. (2010). Environmental risk assessment of offshore produced water discharges using a hybrid fuzzy-stochastic modeling approach. Environmental Modelling & Software, 25, 782–792. Cho, H.-N., Choi, H.-H., & Kim, Y.-B. (2002). A risk assessment methodology for incorporating uncertainties using fuzzy concepts. Reliability Engineering and System Safety, 78, 173–183. Chongfu, H. (1996). Fuzzy risk assessment of urban natural hazards. Fuzzy Sets and Systems, 83, 271–282. Ciaramella, A., Tagliaferri, R., & Pedrycz, W. (2004). Inference systems by using ordinal sums and genetic algorithms. In IEEE annual meeting of the fuzzy information, 2004. Processing NAFIPS ’04 (Vol. 2, pp. 629–634). IEEE. Ciaramella, A., Tagliaferri, R., & Pedrycz, W. (2005). The genetic development of ordinal sums. Fuzzy Sets and Systems, 151, 303–325. Cingolani & Alcala-Fdez, J. (2012). jFuzzyLogic: A robust and flexible Fuzzy-logic inference system language implementation. In Proceedings of IEEE world congress on computational intelligence 2012 (pp. 1–8). IEEE. Cormen, T., Leiserson, C., Rivest, R., & Stein, C. (2009). Introduction to algorithms. The MIT Press. Darbra, R., Demichela, M., & Mure, S. (2008). Preliminary risk assessment of ecotoxic substances accidental releases in major risk installations through fuzzy logic. Process Safety and Environment Protection, 86, 103–111. Davidson, V., Ryks, J., & Fazil, A. (2006). Fuzzy risk assessment tool for microbial hazards in food systems. Fuzzy Sets and Systems, 157, 1201–1210. Dikmen, I., Birgonul, M., & Han, S. (2007). Using fuzzy risk assessment to rate cost overrun risk in international construction projects. International Journal of Project Management, 25, 494–505. Elsayed, T. (2009). Fuzzy inference system for the risk assessment of liquefied natural gas carriers during loading/offloading at terminals. Applied Ocean Research, 31, 179–185. Guimara, A., & Lapa, C. (2007). Fuzzy inference to risk assessment on nuclear engineering systems. Applied Soft Computing, 7, 17–28. Hadjimichael, M. (2009). A fuzzy expert system for aviation risk assessment. Expert System with Application, 36, 6512–6519. Harris, T. (2013). Quantitative credit risk assessment using support vector machine: Broad versus narrow default definitions. Expert Systems with Applications, 40, 4404–4413. Huang, G., Chen, Z., Tontiwachwuthikul, P., & Chakma, A. (2010). Environmental risk assessment for underground storage tanks through an interval parameter fuzzy relation analysis approach. Energy Sources, 21, 75–96. IEC (2000). International Electrotechnical Commission technical committee industrial process measurement and control. IEC 61131 – Programmable controllers. Part 7: fuzzy control programming., Technical report, International electrotechnical commission. Kahraman, C., & Kaya, I. (2009). Fuzzy process accuracy index to evaluate risk assessment of drought effects in turkey. Human and Ecological Risk Assessment, 15, 789–810. Karimi, I., & Hullermeier, E. (2007). Risk assessment system of natural hazards: A new approach based on fuzzy probability. Fuzzy Sets and Systems, 158, 987–999. Lang, A., Brunzel, S., Dolek, M., Otto, M., & Thei, B. (2011). Modelling in the light of key parameters: A call to exercise caution in field prediction of Bt-maize effects. Proceedings of the Royal Society B, Biological Sciences, 278, 980–981.
444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504
Please cite this article in press as: Camastra, F., et al. A fuzzy decision system for genetically modified plant environmental risk assessment using Mamdani inference. Expert Systems with Applications (2014), http://dx.doi.org/10.1016/j.eswa.2014.09.041
430 431 432 433 434 435
438 439 440 441 442
ESWA 9580
No. of Pages 7, Model 5G
15 October 2014 F. Camastra et al. / Expert Systems with Applications xxx (2014) xxx–xxx 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540
Lener, M., Giovannelli, V., Arpaia, S., Burgio, G., Canfora, L., Dinelli, G., et al. (2013). Applying an operating model for the environmental risk assessment in italian sites of community importance (sci) of the european commission habitats directive (92/43/EEC). Bulletin of Insectology, 66, 257–267. Li, J., Huang, G., Maqsood, I., & Huang, Y. (2007). An integrated fuzzy-stochastic modeling approach for risk assessment of groundwater contamination. Journal of Environmental Management, 82, 173–188. Lin, C.-T., & Lee, C. (1996). Neural fuzzy systems: A neuro-fuzzy synergism to intelligent systems. Prentice Hall. Liu, J., Yang, J.-B., Wang, J., & Sii, H.-S. (2005). Engineering system safety analysis and synthesis using the fuzzy rule-based evidential reasoning approach. Quality and Reliability Engineering International, 21, 387–411. Liu, J., Yang, J.-B., Wang, J., Sii, H.-S., & Wang, Y.-M. (2004). Fuzzy rule-based evidential reasoning approach for safety analysis. International Journal of General Systems, 33, 183–204. Li, W., Zhou, J., Xie, K., & Xiong, X. (2008). Power system risk assessment using a hybrid method of fuzzy set and monte carlo simulation. IEEE Transactions on Power Systems, 23. Mokhtari, K., Ren, J., Roberts, C., & Wang, J. (2012). Decision support framework for risk management on sea ports and terminals using fuzzy set theory and evidential reasoning approach. Expert Systems with Applications, 39, 5087–5103. Nedjah, N., & de Macedo Mourelle, L. (2006). Fuzzy systems engineering. Springer. Ngai, E., & Wat, F. (2003). Design and development of a fuzzy expert system for hotel selection. Omega, 31, 275–286. Oreski, S., & Oreski, G. (2014). Genetic algorithm-based heuristic for feature selection in credit risk assessment. Expert Systems with Applications, 41, 951–961. Parr, T., & Quong, R. (1995). ANTLR: A predicated-ll(k) parser generator. Software: Practice and Experience, 25, 789–810. Perry, J., Devos, Y., Arpaia, S., Bartsch, D., Gathmann, A., Hails, R., et al. (2010). A mathematical model of exposure of non-target lepidoptera to Bt-maize pollen expressing Cry1Ab within Europe. Proceedings of the Royal Society B, Biological Sciences, 277, 1417–1425. Petrovic, D., Tanasijevic, M., Milic, V., Lilic, N., Stojadinovic, S., & Svrkota, S. (2014). Risk assessment model of mining equipment failure based on fuzzy logic. Expert Systems with Applications, 41, 8157–8164.
7
Sadiq, R., & Husain, T. (2005). A fuzzy-based methodology for an aggregative environmental risk assessment: A case study of drilling waste. Environmental Modelling & Software, 20, 33–46. Samantra, C., Datta, S., & Shankar Manapantra, S. (2014). Risk assessment in it outsourcing using fuzzy decision-making approach: An indian perspective. Expert Systems with Applications, 41, 4010–4022. Smith, E., & Eloff, J. (2000). Cognitive fuzzy modeling for enhanced risk assessment in a health care institution. IEEE Intelligent Systems and Their Applications, 15, 69–75. Sorlini, C., Buiatti, M., Burgio, G., Cellini, F., Giovannelli, V., & Lener, M., et al. (2013). La valutazione del rischio ambientale dell’ immissione deliberata nell’ ambiente di organismi geneticamente modificati (in Italian), Technical report, Italian ministry of environment, http://bch.minambiente.it/EN/Biosafety/propmet.asp, 2003. Urbanowicz, R., & Moore, J. (2009). Learning classifier systems: A complete introduction, review, and roadmap. Journal of Artificial Evolution and Applications, 2009, 125. Wang, F., Ding, L., Luo, H., & Love, P. (2014). Probabilistic risk assessment of tunneling-induced damage to existing properties. Expert Systems with Applications, 41, 951–961. Wang, Y.-M., & Elhag, T. (2008). An adaptive neuro-fuzzy inference system for bridge risk assessment. Expert Systems with Applications, 34, 3099–3106. Yang, Z., Bonsall, S., & Wang, J. (2008). Fuzzy rule-based bayesian reasoning approach for prioritization of failures in FMEA. IEEE Transactions on Reliability, 57, 517–528. Yang, J.-B., Liu, J., Wang, J., Sii, H.-S., & Wang, H.-W. (2006). Belief rule-base inference methodology using the evidential reasoning approach – RIMER. IEEE Transactions on Systems, Man, and Cybernetics – Part A: Systems and Humans, 26, 266–285. Zeng, J., An, M., & Smith, N. (2007). Application of a fuzzy based decision making methodology to construction project risk assessment. International Journal of Project Management, 25, 589–600. Zhou, J. (2010). Spa-fuzzy method based real-time risk assessment for major hazard installations storing flammable gas. Safety Science, 48, 819–822.
Please cite this article in press as: Camastra, F., et al. A fuzzy decision system for genetically modified plant environmental risk assessment using Mamdani inference. Expert Systems with Applications (2014), http://dx.doi.org/10.1016/j.eswa.2014.09.041
541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575