PROCEEDINGS OF WORLD ACADEMY OF SCIENCE, ENGINEERING AND TECHNOLOGY VOLUME 13 MAY 2006 ISSN 1307-6884

Association Rule and Decision Tree based Methods for Fuzzy Rule Base Generation
Ferenc Peter Pach and Janos Abonyi Pannon University, Department of Process Engineering, Veszprem, P.O. Box 158, H-8201, Hungary, http://www.fmt.vein.hu/softcomp abonyij@fmt.vein.hu

Abstract This paper focuses on the data-driven generation of fuzzy IF...THEN rules. The resulted fuzzy rule base can be applied to build a classier, a model used for prediction, or it can be applied to form a decision support system. Among the wide range of possible approaches, the decision tree and the association rule based algorithms are overviewed, and two new approaches are presented based on the a priori fuzzy clustering based partitioning of the continuous input variables. An application study is also presented, where the developed methods are tested on the well known Wisconsin Breast Cancer classication problem.

I. I NTRODUCTION Human logic can be represented well by logical expressions in syntax of rules, with an antecedent and a consequent part. A short example can be: If somebody has forgotten her/his umbrella at home and it is pouring with rain then the chances are that she/he will be ooding. The set of logical rules is called rule base that is an easy and useful interpretation of the knowledge of a given area. Various types of logical rules can be discussed in the context of the decision borders these rules create in multidimensional feature space. The standard crisp propositional IF...THEN rules provide overlapping hyperrectangular covering areas, threshold logic rules are equivalent to separating hyperplanes, while fuzzy rules based on real-valued predicate functions (come from the prolog to [52]). Accordingly many rule based methods have been developed for extraction knowledge from databases. The paper [40] introduces a genetic programming (GP) and fuzzy logic based algorithm that extracts explanatory rules from micro array data. A hybrid approach is proposed in [7], where a standard GP and a heuristic hierarchical crisp rule-base construction are combined. A fuzzy mining algorithm based on Srikant and Agrawals method [48] is proposed for extracting generalized rules with the use of taxonomies [51]. In [34] compact fuzzy rules extraction is based on adaptive data approximation using B-splines. Rule bases are efciently used in many area but this paper concentrates rst of all to the prediction applications. Rule bases are successfully applied for example in stock exchange estimation [37], weather [32] or future sales forecasting [19]. The high prediction accuracy of the applied model (build from the extracted rules) is very important but the model understanding could be also very critical in many areas. It

is very useful to know what are in the background of the decisions, while rules could be edited or changed by the specialists of the application area. The compact and apprehensible predictive models via the visualization possibilities could help better human decisions. The paper [52] shows many computational intelligence techniques (based on decision trees, neural networks, etc.) that very useful tools to rule extraction and data understanding. In developments of the new rule based methods for prediction applications besides the retention and enhancement of achieved accuracies (in the classication problems), the one of the most important objects is to enlarge the interpretable of the rules. To take this aspect into account the one of the possible improvement ways is the adaptation of fuzzy logic. Besides the fuzzy methods could represent the discovered rules far natural for human, the fuzzy logic serves more robust predictive models (classiers) in case of false, inconsistent, and missing data. In this paper a fuzzy decision tree (Section II-B) and a fuzzy association rule based method (Section III-B) are introduced for fuzzy rule base generation. Our main goal is to show how construct compact fuzzy rule bases which can be used for data analysis, classication, or prediction. Therefore prediction accuracy (for classication problems) and understanding are together in focus during the rule extraction steps in both algorithms. The classication effectiveness of the proposed methods are tested on the Wisconsin Breast Cancer problem. The results are summarized in a short application study (Section IV). II. F UZZY DECISION TREE BASED METHODS A. Existent decision tree induction algorithms Decision tree based methods are widely used in data mining and decision support applications. Decision tree is fast and easy to use for rule generation and classication problems, moreover it is an excellent representation tool of decisions. The popularity and the spread of decision tree are based on the algorithm ID3 by Quinlan [46]. Many studies had been written to induction and analysis of decision trees [54], [47], [35], [36], [55]. The application areas of decision trees are also very breadth [6], [45], [15], [50], [49], [38].

PWASET VOLUME 13 MAY 2006 ISSN 1307-6884

45

2006 WASET.ORG

PROCEEDINGS OF WORLD ACADEMY OF SCIENCE, ENGINEERING AND TECHNOLOGY VOLUME 13 MAY 2006 ISSN 1307-6884

Bare nuclei
small large

Data Partition method Prepruning Decision tree induction Fuzzy rule base Postpruning

Data Partition method Frequent itemsets Rule generation Fuzzy rule base Postpruning

Cell size
small large

Malignant

Benign

Cell shape
large

Malignant

Fig. 1. problem

Fuzzy decision tree for Wisconsin Breast Cancer classication

Since the 80s years many fuzzy decision tree induction algorithm have been introduced [2], [42], [39]. Fuzzy decision trees represent the discovered rules far natural for human (for example thanks to the linguistic variables). The [22] takes a detailed introduction about the non fuzzy rules and the different kind of fuzzy rules. The Figure 1 shows an example fuzzy decision tree for a classication problem. The aim of the classication is to distinguish between benign and malignant cancers based on the available attributes. The example tree uses only three attributes (three decision points: bare nuclei, cell size and cell shape) and represents three rules (the pathes from root to the letters) to the decision. In classication problems the continuous attributes in the input domain need partitioning. For example in Figure 1 the attribute cell size is partitioned into two overlapped partitions (two fuzzy sets) small and large. Many type of membership functions can be used (triangular, trapezoids, Gaussian, etc.) for partitions. While the papers in the literature discuss various methods, this paper focuses only the a priori partition based fuzzy decision tree induction algorithms. At the a priori based methods, the partition step is ahead the tree induction step. A new a priori partition and decision tree based extraction method is showed in the next subsection. B. A fuzzy decision tree based method Our method (Figure 2, on the left) consists the following main steps: 1) A supervised clustering algorithm is used for input domain partition. The supervised method takes into account the class labels during the clustering. Therefore the resulted partitions, the fuzzy membership functions (fuzzy sets) represent not only the distribution of data, but the distribution of the classes too. 2) During a pre-pruning method the resulted partitions

Fig. 2. Main steps of the decision tree (left) and the association rule (right) based methods

could be analyze and combine the unduly overlapped fuzzy sets. 3) The results of the pre-pruning step are input parameters (beside data) for the tree induction algorithm. The applied tree induction method is the FID (Fuzzy Induction on Decision Tree) algorithm by C. Z. Janikow [35]. 4) The resulted decision tree is analyzed and transformed by a proper method into a fuzzy rule base. 5) While the FID algorithm could generate more large and complex decision tree as it is necessary, therefore a postpruning method is applied to lter the unnecessary long rules and erase the weak (in classication point of view) rules from the fuzzy rule base. This method provides compact and transparent fuzzy rule base which can be use to build accurate fuzzy classiers. III. F UZZY ASSOCIATION RULE BASED METHODS A. Existent association rule mining and associative classier algorithms Besides the decision tree based techniques the association rule mining algorithms are the most frequently used data mining tools in rule extraction. Many kinds of methods are developed [3], [5], [4], [16], [18], [8], [12], [13], [14], [9], [11], [21], [17], [20] but two main steps are common in most of them. The mining starts with frequent item set searching (it is dened rst in paper [3]) then association rules are generated from the large item sets. The selection of an appropriate algorithm depends on the structure (sparse, dense) and the size of the analyzed database. Additionally the application area inuences also notable the suitable methods. The rst association rule mining algorithms primarily developed to discover the customer habits in the market basket analysis [5]. See an example transactional database in Table I. All the

PWASET VOLUME 13 MAY 2006 ISSN 1307-6884

46

2006 WASET.ORG

PROCEEDINGS OF WORLD ACADEMY OF SCIENCE, ENGINEERING AND TECHNOLOGY VOLUME 13 MAY 2006 ISSN 1307-6884

T 1 2 3 4 5

Products milk, bread, beer, egg bread, diaper, egg, beer milk, bread milk, diaper, wine, beer milk, bread, diaper, beer TABLE I

500 450 400 350

E XAMPLE TRANSACTIONAL DATABASE

300 250 200 Benign Malignant

rows have a transaction identity number (T) and each rows contain products buy together in the transaction. The aim is to understand the behavior of retail customers, or in other words, nding associations among the items purchased together. The products are called items and the item sets are the sets of the products. An item (item set) is called frequent item (item set) if it has higher support (the number of the occurences in database the number of it is purchased) as the predened minimum support threshold. For example if the minimal support is set to fty percentage, the item set <diapers, beer> is a frequent item set. A famous example of an association rule in such a database is diapers => beer, i.e. young fathers being sent off to the store to buy diapers, reward themselves for their trouble. An association rule have a condence measure that represents the strength of the relationship between the antecedent and consequent parts of the rule. An association rule is called valid rule if and only if the support and condent values are higher than the support and condence thresholds. Besides the possibility of the rule based analysis of the transactional databases, classiers can also built from the set of discovered association rules. The CBA algorithm [43] was the rst which integrates efcient the association and classication rule mining techniques. In last decade many associative classier algorithms are presented [23], [44], [53], [41], [57], [56], [58], [10], [59], [30]. The methods give rule bases with higher and higher classication power, but the most of them generate too large and complex classiers. How it has been already accentuated in Section I and Section II too complex rule bases are undesirable in aspect of the interpretability. Therefore our main goal was to construct an associative method which serves compact fuzzy rule bases from data which is applicable to build accurate fuzzy classiers. The next subsection introduces our new fuzzy association rule based method. B. A fuzzy association rule based method Our method (Figure 2, on the right) has the following main steps: 1) In the rst step a partitioning method is need to get discrete data elements on continuous attributes. The applied method is a fuzzy clustering algorithm to determine trapezoidal fuzzy membership functions for each attributes. 2) While the membership functions as fuzzy sets are counted for fuzzy items, the frequent item sets are searched on easy way. The membership values determines the supports of the items. The searching of the

Fig. 3.

Class distribution of Wisconsin Breast Cancer Data

larger item sets is based on the Apriori-principle [5]. 3) While our main application goal is the classier model identication, association rules with class label in the consequent part must be generated from the frequent item sets. 4) The classication rules determine most the results of prediction are selected by a correlation measure. These rules are called important rules. Only the positive correlated, above the average rules are stored in the rule base. The proposed method efciently works without database coverage analysis (which demands high computational capacity). 5) The unnecessary complex, redundant and conict rules are searched during a post-pruning method. The selected rules are removed from the rule base therefore only the most important and most condential rules could be use for fuzzy associative classiers. The earlier versions of both presented methods are detailed in our publications [29], [24], [25], [27], [26], [28]. Our actual results are encouraging, the classication power and complexity reduction of the presented methods are demonstrated with a short application study in the following section. IV. A PPLICATION STUDY If only classication rules is generated by the proposed methods, the rule bases and the input partitions serve classication models. This section shows an empirical analysis of the classication power of the proposed algorithms. The Wisconsin Breast Cancer data (WBCD) is available from the University of California, Irvine (UCI Repository, http://www.ics.uci.edu/ mlearn/MLRepository.html), is a real classication problem. The aim of the classication is to distinguish between benign and malignant cancers based on the available nine measurements: clump thickness, uniformity of cell size, uniformity of cell shape, marginal adhesion, single epithelial cell size, bare nuclei, bland chromatin, normal nuclei, and mitosis. The attributes have integer value in the range [1;10]. The original database contains 699 instances however 16 of these are omitted because these are incomplete, which is common with other studies. The class distribution (Figure 3) is 65.5 benign and 34.5 malignant, respectively.

PWASET VOLUME 13 MAY 2006 ISSN 1307-6884

47

2006 WASET.ORG

PROCEEDINGS OF WORLD ACADEMY OF SCIENCE, ENGINEERING AND TECHNOLOGY VOLUME 13 MAY 2006 ISSN 1307-6884

1 0.5 0 1 0.5 0 1 0.5 0 1 0.5 0 1 0.5 0

thickness

1 0.5

cell size

thickn. c. size

c. shape

adhes. e.c. size b.nuclei

chrom.

n. nuclei

mitosis b/m
0.8638

0.2

0.4

0.6

0.8

0 1 0.5 0 1 0.5 0 1 0.5

0.2

0.4

0.6

0.8

cell shape

adhesion

0.825

0.2

0.4

0.6

0.8

0.2

0.4

0.6

0.8

0.7775

ep.cell size

b.nuclei
0.7732
0 0.2 0.4 0.6 0.8 1

0.2

0.4

0.6

0.8

0.7658

chromatin

n. nuclei
0.7646

0.2

0.4

0.6

0.8

0.2

0.4

0.6

0.8

mitosis

0.7515

0.2

0.4

0.6

0.8

0.7362

Fig. 4. Partitions (fuzzy trapezoidal membership functions) determined by supervised Gath-Geva clustering algorithm for Wisconsin classication problem)
4 7 3 7 3 7 3 7 3 6 3 8 3 7 3 7 2 7

0.7225

0.695

A. Classication by the fuzzy decision tree based method First see the results of our decision tree based algorithm. The selected a priori partition method was the supervised Gath-Geva clustering algorithm [31], [1]. The number of the initial number of the partitions for all attributes were equal with the number of classes, two. The resulted partitions are represented in Figure 4. The classication accuracy is measured by ten-fold cross validation. If the post-pruning factor is set to 1.6 (in the fth step of our algorithm), the average accuracy is 95.27% with 3.2 rules (number of the conditions: 6.8). An example rule base contains three fuzzy rules is the following: If uniformity of cell size is small and bare nuclei is small Then benign. If uniformity of cell size is large and uniformity of cell shape is large and bare nuclei is small Then malignant. If bare nuclei is large Then malignant. The decision tree contains the rule base is represented in Figure 1. It is a very compact, interpretable, but accurate fuzzy classication rule base for the Wisconsin problem. B. Classication by the fuzzy association rule based method First in the association rule based methods an implementation of the Gustafson-Kessel (GK) clustering algorithm is applied to partition the input attributes [33]. As it was in case of the decision tree based method, the number of the partitions for all attributes were two. The average (by ten-fold cross validation) classication accuracy is 95.85%. A visualization tool is also developed to represent the resulted fuzzy rule base structure. The Figure 5 represents an example rule base contains ten rules with 22.7 conditions. If the GK algorithm is changed with the easiest technique the Ruspini-type fuzzy partition method, more accurate (96%) and smaller (average 8.8 rules), but more complex (average 36.8 conditions) classier is resulted. An example rule base is depicted in Figure 6. Some rules are contained in both rule bases of methods but the gures represents that the associative method serves larger

Fig. 5. Fuzzy rule base for Wisconsin Breast Cancer classication problem (is generated by the association based method with Gustafson-Kessel clustering algorithm as partition technique)

thickn. c. size c. shape adhes. e.c. size b.nuclei chrom.

n. nuclei

mitosis

b/m
0.6556

0.6395

0.6303

0.5948

0.5824

0.5563

0.5421

0.5412

0.5371 1 10 1 10 1 10 1 10 1 10 0 10 1 10 1 10 1 10

Fig. 6. Fuzzy rule base for Wisconsin Breast Cancer classication problem (is generated by the association based method with Ruspini-type partition technique)

rule bases (in both partition techniques) as the decision tree based algorithm. In Figure 5 the fth rule is equal with the third rule in the tree (If bare nuclei is large Then malignant). But rule base of the decision tree includes more compact the important knowledge to classication. See for example the rules number eight and nine together appear in the second rule of the tree.

PWASET VOLUME 13 MAY 2006 ISSN 1307-6884

48

2006 WASET.ORG

PROCEEDINGS OF WORLD ACADEMY OF SCIENCE, ENGINEERING AND TECHNOLOGY VOLUME 13 MAY 2006 ISSN 1307-6884

V. C ONCLUSIONS This paper gave a short overview of the existent decision tree and association rule mining based rule extraction methods focused to build a fuzzy classier system. Beside the literature review, two new rule extraction methods have been presented to generate compact and accurate fuzzy rule base classiers. The results show the similarities of the two approaches, and highlight that the partitioning of the input variables plays an important role to the performance of the resulted classiers. R EFERENCES
[1] J. Abonyi, B. Feil, S. Nemeth, and P. Arva. Modied Gath-Geva clustering for fuzzy segmentation of multivariate time-series. Fuzzy Sets and Systems, Data Mining Special Issue, pages in print, avaiable online from Science Direct, 2005. [2] J. M. Adamo. Fuzzy decision trees. Fuzzy Sets and Systems, 4(3):207 219, 1980. [3] R. Agrawal, T. Imielinski, and A. Swami. Mining association rules between sets of items in large databases. In Proceedings of the 1993 ACM SIGMOD International Conference on Management of Data, pages 207216, 1993. [4] R. Agrawal, H. Mannila, R. Srikant, H. Toivonen, and A.I. Verkamo. Fast discovery of association rules. In Advances in Knowledge Discovery and Data Mining, pages 307328. AAAI/MIT Press, 1996. [5] R. Agrawal and R. Srikant. Fast algorithm for mining association rules in large databases. In Proceedings of the 20th International Conference on Very Large Data Bases, pages 487499, 1994. [6] Jean-Roger Le Gall Anke Neumann, Josiane Holstein and Eric Lepage. Measuring performance in health care: case-mix adjustment by boosted decision trees. Articial Intelligence in Medicine, 32(2):97113, 2004. [7] Jan Jantzen Hubertus Axer Beth Bjerregaard Athanasios Tsakonas, Georgios Dounias and Diedrich Graf von Keyserlingk. Evolving rulebased systems in two medical domains using genetic programming. Articial Intelligence in Medicine, 32(3):195216, 2004. [8] W-H. Au and K.C.C. Chan. An effective algorithm for discovering fuzzy rules in relational databases. In Proceedings of the 7th IEEE International Conference on Fuzzy Systems, pages 13141319, 1998. [9] W-H. Au and K.C.C. Chan. Farm: A data mining system for discovering fuzzy association rules. In Proceedings of the 8th IEEE International Conference on Fuzzy Systems, pages 12171222, 1999. [10] Elena Baralis and Silvia Chiusano. Essential classication rule sets. ACM Transactions on Database Systems, 29(4):635674, 2004. [11] Y. Bastide, R. Taouil, N. Pasquier, G. Stumme, and L. Lakhal. Mining frequent patterns with counting inference. SIGKDD Explorations, 2(2):6675, 2000. [12] R.J. Bayardo. Efciently mining long patterns from databases. In Proceedings of the 1998 ACM SIGMOD International Conference on Management of Data, pages 8593, 1998. [13] R.J. Bayardo and R. Agrawal. Mining the most interesting rules. In Proceedings of the 1999 ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pages 145154, 1999. [14] R.J. Bayardo, R. Agrawal, and D. Gunopulos. Constraint-based rule mining in large, dense databases. In Proceedings of the 1999 IEEE International Conference on Data Engineering, pages 188197, 1999. [15] C.J. Moran B.L. Henderson, E.N. Bui and D.A.P. Simon. Australiawide predictions of soil properties using decision trees. Geoderma, 124(3-4):383398, 2005. [16] S. Brin, R. Motwani, J. Ullman, and S. Tsur. Dynamic itemset counting and implication rules for market basket data. In Proceedings of the 1997 ACM SIGMOD International Conference on Management of Data, pages 255264, 1997. [17] D. Burdick, M. Calimlim, and J. Gehrke. Maa: A maximal frequent itemset algorithm for transactional databases. In Proceedings of the 2001 IEEE International Conference on Data Engineering, pages 443552, 2001. [18] K.C.C. Chan and W-H. Au. Mining fuzzy association rules. In Proceedings of the 1997 International Conference on Information and Knowledge Management, pages 209215, 1997.

[19] Liu C-H. Wang Y-W. Chang, P-C. A hybrid model by clustering and evolving fuzzy rules for sales decision supports in printed circuit board industry. Decisions Support Systems, Available online 13 December 2005. [20] G. Chen and Q. Wei. Fuzzy association rules and the extended mining algorithms. Information Sciences, 147:201228, 2002. [21] G. Chen, Q. Wei, and E. Kerre. Fuzzy data mining: Discovery of fuzzy generalized association rules. In Recent Issues on Fuzzy Databases, pages 4566. Springer, 2000. [22] Henri Prade Didier Dubois. What are fuzzy rules and how to use them. Fuzzy Sets and Systems, 84:169185, 1996. [23] Zhang X. Wong-L. Li J. Dong, G. Caep: classication by aggregating emerging patterns. In Second International Conference on Discovery Science, 1999. [24] J. Abonyi F. D. Tamas, F. P. Pach and A. M. Esteves. Analysisof trace elements in clinker based on supervised clustering and fuzzy decision tree induction. In 6th International Congress, Global Construction: Ultimate Concrete Opportunities, Dundee, Scotland, 2005. [25] P. Arva F. P. Pach, A. Gyenesei and J. Abonyi. Fuzzy association rule mining for model structure identication. In 10th Online World Conference on Soft Computing in Industrial Application, WSC10, 2005. [26] P. Arva F. P. Pach, A. Gyenesei and J. Abonyi. Fuzzy association rule mining for model structure identication. In Applications of Soft Computing: Recent Trends, Springer, 2006, In Press. [27] S. Nemeth P. Arva J. Abonyi F. P. Pach, A. Gyenesei. Fuzzy association rule mining for the analysis of historical process data. Acta Agraria Kaposvariensis, 2006, In Press. [28] S. Nemeth P. Arva J. Abonyi F. P. Pach, F. Szeifert. Fuzzy association rule mining for data-driven analysis of dynamical systems. Hungarian Journal of Industrial Chemistry, Special Issue on Recent advances in Computer Aided Process Engineering, 2006, In Press. [29] S. Nemeth P. Arva F.P. Pach, J. Abonyi. Supervised clustering and fuzzy decision tree induction for the identication of compact classiers. In 5th International Symposium of Hungarian Researchers on Computational Intelligence, Budapest, Hungary, 2004. [30] Paul Leng Frans Coenen. The effect of threshold values on association rule based classication accuracy. Data and Knowledge Engineering, Available online, 2006. [31] I. Gath and A.B. Geva. Unsupervised optimal fuzzy clustering. IEEE Transactions on Pattern Analysis and Machine Intelligence, 7:773781, 1989. [32] Pal N.R. Das J. Ghosh, A. A fuzzy rule based approach to cloud cover estimation. Remote Sensing of Environment, 100:531549, 2006. [33] D.E. Gustafson and W.C. Kessel. Fuzzy clustering with fuzzy covariance matrix. In In Proceedings of the IEEE CDC, San Diego, pages 761766, 1979. [34] S. Kper J. Zhang and A. Knoll. Extracting compact fuzzy rules based on adaptive data approximation using b-splines. Information Sciences, 142(1-4):227248, 2002. [35] C.Z. Janikow. Fuzzy decision trees: issues and methods. IEEE Trans. Systems Man Cybernet. Part B (Cybernetics), 28(1):114, 1998. [36] C.Z. Janikow. Fuzzy partitioning with d 3.1. In Proc. 18th Internat. Conf. of the North American Fuzzy Information Processing Society, NAFIPS99, pages 467471, 1999. [37] Patrick Soriano Jean-Yves Potvin and Maxime Valle. Generating trading rules on the stock markets with genetic programming. Computers and Operations Research, 31(7):10331047, 2004. [38] E.S. Karapidakis. Machine learning for frequency estimation of power systems. Applied Soft Computing, In Press, Corrected Proof, Available online 28 December 2005. [39] Kun Chang Lee and Sung Joo Park. A knowledge-based fuzzy decision tree classier for time series modeling. Fuzzy Sets and Systems, 33(1):1 18, 1989. [40] Ricardo Linden and Amit Bhaya. Evolving fuzzy rules to model gene expression. Biosystems, In Press, Accepted Manuscript, Available online 30 April 2006. [41] Ma Y. Liu, B. and Wong C. K. Improving an association rule based classier. In Principles of Data Mining and Knowledge Discovery, pages 504509, 2000. [42] T. Bar-Noy M. Friedman and M. Blau A. Kandel. Certain computational aspects of fuzzy decision trees. Fuzzy Sets and Systems, 28(2):163170, 1988. [43] Bing Liu Wynne Hsu Yiming Ma. Integrating classication and association rule mining. In Appeared in KDD-98, New York, 1998.

PWASET VOLUME 13 MAY 2006 ISSN 1307-6884

49

2006 WASET.ORG

PROCEEDINGS OF WORLD ACADEMY OF SCIENCE, ENGINEERING AND TECHNOLOGY VOLUME 13 MAY 2006 ISSN 1307-6884

[44] D. Meretakis and B. Wuthrich. Extending naive bayes classiers using long itemsets. In Knowledge Discovery and Data Mining, pages 165 174, 1999. [45] A. Keith Dunker Predrag Radivojac, Nitesh V. Chawla and Zoran Obradovic. Classication and knowledge discovery in protein databases. Journal of Biomedical Informatics, 37(4):224239, 2004. [46] J. R. Quinlan. Induction on decision trees. Machine Learning, 1(1):81 106, 1986. [47] J.R. Quinlan. C4.5: Programs for Machine Learning. Morgan Kaufmann, San Mateo, CA, 1993. [48] R. Agrawal R. Srikant. Mining generalized association rules. In The Internat. Conf. on Very Large Databases, 1995. [49] A. Abuelgasim R.H. Fraser and R. Latifovic. A method for detecting large-scale forest cover change using coarse spatial resolution imagery. Remote Sensing of Environment, 95(4):414427, 2005. [50] Sbastien Thomassey and Antonio Fiordaliso. A hybrid sales forecasting system based on clustering and decision trees. Decision Support Systems, In Press, Corrected Proof,, Available online 30 March 2005. [51] Kuei-Ying Lin Tzung-Pei Hong and Shyue-Liang Wang. Fuzzy data mining for interesting generalized association rules. Fuzzy Sets and Systems, 138(2):255269, 2003. [52] J. M. Zurada W. Duch, R. Setiono. Computational intelligence methods for rule-based data understanding. Proc. of the IEEE, 92(5), 2004. [53] Zhou S. Wang, K. and Y. He. Growing decision tree on support-less association rules. In In proceedings of KDD00, Boston, MA, 2000. [54] R. Weber. Fuzzy id3: a class of methods for automatic knowledge acquisition. In Proc. 2nd Internat. Conf. on Fuzzy Logic and Neural Networks, Iizuka, Japan, page 265268, 1992. [55] Zenon A. Sosnowskic Witold Pedrycz. The design of decision trees in the framework of granular data and their application to software quality models. Fuzzy Sets and Systems, 123:271290, 2001. [56] Gwo-Hshiung Tzeng Yi-Chung Hu. Elicitation of classication rules by fuzzy data mining. Engineering Applications of Articial Intelligence, 16:709716, 2003. [57] Gwo-Hshiung Tzeng Yi-Chung Hu, Ruey-Shun Chen. Mining fuzzy association rules for classication problems. Computers and Industrial Engineering, 43:735750, 2002. [58] X. Yin and J. Han. Cpar: Classication based on predictive association rules. In in Proceedings of 2003 SIAM International Conference on Data Mining (SDM03), 2003. [59] A. Zimmermann and Raedt L. D. Corclass: Correlated association rule mining for classication. In Discovery Science, 7th International Conference, Padova, Italy, pages 6072, 2004.

PWASET VOLUME 13 MAY 2006 ISSN 1307-6884

50

2006 WASET.ORG