![David C. Wyld et al. (Eds) : SAI, CDKP, ICAITA, NeCoM, SEAS, CMCA, ASUC, Signal - 2014
pp. 105–114, 2014. © CS & IT-CSCP 2014 DOI : 10.5121/csit.2014.41111
A BOOLEAN MODELING FOR IMPROVING
THE ALGORITHM APRIORI
Abdelhak Mansoul1
and Baghdad Atmani2
1
Computer Science Laboratory of Oran (LIO),
Department of Computer Sciences, University of Skikda, Algeria.
mansoul.abdelhak@yahoo.fr
2
Computer Science Laboratory of Oran (LIO),
Department of Computer Science, University of Oran ES-Sénia, Algeria.
atmani.baghdad@gmail.com
ABSTRACT
Mining association rules is one of the most important data mining tasks. Its purpose is to
generate intelligible relations between attributes in a database. However, its use in practice is
difficult and still raises several challenges, in particular, the number of learned rules is often
very large. Several techniques for reducing the number of rules have been proposed as
measures of quality, syntactic filtering constraints, etc. However, these techniques do not limit
the shortcomings of these methods. In this paper, we propose a new approach to mine
association, assisted by a Boolean modeling of results in order to mitigate the shortcomings
mentioned above and propose a cellular automaton based on a boolean process for mining,
optimizing, managing and representing of the learned rules.
KEYWORDS
Cellular automaton, Data mining, Association Rules, Boolean modeling, Apriori-Cell
1. INTRODUCTION
Numerous studies on the association rules are made [2], [9], [11]. However, their uses in practice
are difficult and still raises many challenges, especially the exorbitant number of rules learned,
and the processing time. Recent studies have also proposed a series of solutions to improve the
performance of the mining process [4], [5], [7], [15], [16], without eliminating the shortcomings
of this method of search data. It became necessary to find adequate techniques and algorithmic
solutions to minimize the cost for space and computing time. The Apriori algorithm introduced
an approach called "test-and-generate" with pruning. However, this approach suffers from a
number of candidates that generates, particularly for relatively small values of support. However,
these approaches do not limit the shortcomings of these methods. Given this situation, it became
necessary to invest in new methods to faces the following challenges:
− Find heuristics to prune the search space;
− Find technical or algorithmic solutions, specifically adequate data structures, to minimize
the cost in space or in process time.](https://image.slidesharecdn.com/abooleanmodelingforimproving-141116210223-conversion-gate02/85/A-boolean-modeling-for-improving-1-320.jpg)
![106 Computer Science & Information Technology (CS & IT)
We will expose in our present article, the second part of our study (see 3.2 Step 4) and its
experimentation that was performed with the basic Apriori in order to demonstrate relevance and
efficiency of the approach that we have considered. Later (continuation of our study) we will
present the first part (see 3.2, Steps 2 and 3), we adopt Apriori-Cell.
2. RELATED WORK
Recent studies have proposed a series of solutions to improve the performance of extraction
process of frequent item sets, including cellular automata [10]. Solutions were oriented
essentially on the Reduction of I / O and the minimization of the cost of the step of computing
the support [9]. Other studies have been based on the discovery of "closed" item sets arising from
the theory of formal concepts [11]. Others propose to generate a representative base or generic
association rules [14] [16] and used techniques to reduce the number of rules with the use quality
measurements [15], syntactic filtering by constraints [2].
3. THE PROPOSED APPROACH
We propose a new approach being located at the junction of two domains that are the Knowledge
Discovery from Databases (KDD) on one hand and representation of knowledge from the other.
Our approach proceeds in three steps:
1. Extraction of frequent patterns and generating association rules using the algorithm
Apriori-Cell which operates on a cell basis;
2. Boolean modelling for association rules;
3. Rules management by the inference engine ICR of the cellular automaton.
A cellular automaton is a grid composed by cells which change states in discreet steps. After
each step, the status of each cell is modified as states of its neighbors in the previous step [1].
Our approach is implemented by two modules:
• The module MAR (Mining Association Rules) ;
• The module ICR (Induction of Cellular Rules).
The dynamics of the cellular automaton. The inference engine of the cellular automaton
simulates the basic operating principle of a classical inference engine using two finite layers of
finite automata. The first layer CEL Fact/CEL Item for the basic facts/Items and the second layer,
CEL Rule/CEL Transaction for the basic Rules/Transactions. Each cell at time t+1 depends only
on the status of its neighbors and his own at time t.](https://image.slidesharecdn.com/abooleanmodelingforimproving-141116210223-conversion-gate02/85/A-boolean-modeling-for-improving-2-320.jpg)
![Computer Science & Information Technology (CS & IT) 113
Table 2. Evolution of execution time (Basic Apriori and global)
Storage space. We find that cell representation is more interesting, and it will be much most
prominently with a more substantial sample.
Figure 4 . Evolution of storage space
6. CONCLUSION
After describing the disadvantages of rule-based methods in data mining, we have proposed an
extraction rules guided by a Boolean modeling based on the Boolean principle of cellular
automata in order to have a base rules optimized and reduced processing time enough, and thus
make a contribution to the construction of knowledge-based systems by adopting a new cellular
technic Thus, the advantages of our method based on the cellular automaton can be summarized
as follows:
• Simple and minimal preprocessing of association rules base, for its transformation into
binary matrix according to the principle of cell layers,
• Ease of implementation functions δfact and δrule that are low complexity and well adapted
to situations with many attributes of rules.
REFERENCES
[1] Atmani, B., Beldjilali, B. (2007) “Knowledge Discovery in Database : induction graph and cellular
automaton”, Computing and Informatics Journal, Vol. 26 N°2 171-197
[2] Besson, J., Robardet, c., Boulicaut, J.F. (2004) “Constraint-based mining of formal concepts in
transactional data”, Conference on Knowledge Discovery and Data Mining (PAKDD’04) volume 3056
of LNCS, Sydney- Australia, pp. 615–624
[3] Pasquier, N., Bastide, Y., Taouil, R., Lakhal, L., Stumme, G. (2000) “Mining minimal non-redundant
association rules using frequent closed itemsets”, Proceedings of the Intl. Conference DOOD’2000,
LNCS, Springer-verlag, July, pp. 972-986
[4] Hajek, P., Havel, I., Chytil, M. (1966) “The GUHA method of automatic hypotheses determination”,
Computing 1, pp. 293-308
Confidence
%
Support
%
Number of
Genes
Generated
items
Number
of rules
Execution
of Apriori
Global
Execution
10 30 12 37 69 0.00 s 0.00 s
50 50 12 37 125874 0.67 s 1.69 s
70 60 12 37 786756 3.56 s 6.17 s](https://image.slidesharecdn.com/abooleanmodelingforimproving-141116210223-conversion-gate02/85/A-boolean-modeling-for-improving-9-320.jpg)
![114 Computer Science & Information Technology (CS & IT)
[5] Boullé, M. (2007) “Recherche d’une représentation des données efficace pour la fouille des grandes
bases de données”, Thèse de Doctorat, Ecole Nationale Supérieure des Télécommunications Paris.
[6] Abdelouhab, F., Atmani, B. (2008) “Intégration automatique des données semi-structurées dans un
entrepôt cellulaire”, Troisième atelier sur les systèmes décisionnels, Mohammadia– Maroc, 10-11
octobre, pp. 109-120
[7] Agrawal, R., Imielinski, T., Swami, A. (1993) “Mining associations between sets of items in large
databases”, Proc. of the ACM SIGMOD Conf., Washington DC, USA
[8] Agrawal, R., Srikant, R. (1994) “Fast Algorithms for Mining Association Rules”, Proceedings of the
20th International Conference on Very Large Data Bases, Santiago, Chile, pp 487-499
[9] Bykowski, A., Rigotti, C. (2001) “A condensed representation to find frequent patterns”, Proceedings
of the Twentieth ACM SIGACTSIGMOD SIGART Symposium on Principles of Database Systems,
ACM Press, pp. 267-273
[10]Fawcett, T. (2008) “ Data mining with cellular automata”, ACM SIGKDD Explorations
Newsletter,v.10 n.1, June 2008.
[11]Ganter, B., Wille, R. (2004) “Conceptual graphs and formal concept analysis”, Lecture Notes in
Computer Science Volume 1257, 1997, pp 290-303Springer-Verlag
[12]Mansoul, A., Atmani, B. (2009) “Fouille de données biologiques : vers une représentation booléenne
des règles d’association”, CEUR-WS:04-Dec-2009/Vol-547, Conférence Internationale sur
l’Informatique et ses Applications CIIA’09, Saida – Algérie
[13]Mansoul, A., Atmani B. (2010) “Vers un automate cellulaire pour la fouille de données : Partie I : la
représentation booléenne des règles d’association”, ASD 5/6 Novembre 2010, Sfax, Tunisie.
[14]Pasquier, N., Bastide Y., N., Lakhal, L. (2000) “Mining minimal non-redundant association rules using
frequent closed itemsets”, Computational logic. International conference No1, London, royaume uni
(24/07/2000), vol. 1861, pp. 972-986.
[15]Vaillant, B., Meyer, P., Prudhomme, E., Lallich, S., Lenca, P., Bigaret, S. (2005) “Mesurer l'intérêt des
règles d'association”, Atelier Qualité des Données et des Connaissances (DQK 05, Actes de EGC), pp.
69-78
[16]Stumme, G., Taouil, R., Bastide, Y., Pasquier, N., Lakhal, L. (2001) “Intelligent structuring and
reducing of association rules with formal concept analysis”, Proc. KI’2001. conference, LNAI 2174,
Springer.
AUTHORS
Abdelhak Mansoul is an Assistant Professor at Skikda University and affiliated researcher in Oran
Computer Lab of Oran University. His research interests are in Database Management System, Data
Mining, decision support systems, and simulation.
Baghdad Atmani is a professor in computer science at the University of Oran (Algeria). His interest field
is Data Mining and Machine Learning Tools. His research is based on Knowledge Representation,
Knowledge-based Systems and CBR, Data and Information Integration and Modeling, Data Mining
Algorithms, Expert Systems and Decision Support Systems.His research are guided and evaluated through
various applications in the field of control systems, scheduling, production, maintenance, information
retrieval, simulation, data integration and spatial data mining.](https://image.slidesharecdn.com/abooleanmodelingforimproving-141116210223-conversion-gate02/85/A-boolean-modeling-for-improving-10-320.jpg)
A boolean modeling for improving
- 1. David C. Wyld et al. (Eds) : SAI, CDKP, ICAITA, NeCoM, SEAS, CMCA, ASUC, Signal - 2014 pp. 105–114, 2014. © CS & IT-CSCP 2014 DOI : 10.5121/csit.2014.41111 A BOOLEAN MODELING FOR IMPROVING THE ALGORITHM APRIORI Abdelhak Mansoul1 and Baghdad Atmani2 1 Computer Science Laboratory of Oran (LIO), Department of Computer Sciences, University of Skikda, Algeria. [email protected] 2 Computer Science Laboratory of Oran (LIO), Department of Computer Science, University of Oran ES-Sénia, Algeria. [email protected] ABSTRACT Mining association rules is one of the most important data mining tasks. Its purpose is to generate intelligible relations between attributes in a database. However, its use in practice is difficult and still raises several challenges, in particular, the number of learned rules is often very large. Several techniques for reducing the number of rules have been proposed as measures of quality, syntactic filtering constraints, etc. However, these techniques do not limit the shortcomings of these methods. In this paper, we propose a new approach to mine association, assisted by a Boolean modeling of results in order to mitigate the shortcomings mentioned above and propose a cellular automaton based on a boolean process for mining, optimizing, managing and representing of the learned rules. KEYWORDS Cellular automaton, Data mining, Association Rules, Boolean modeling, Apriori-Cell 1. INTRODUCTION Numerous studies on the association rules are made [2], [9], [11]. However, their uses in practice are difficult and still raises many challenges, especially the exorbitant number of rules learned, and the processing time. Recent studies have also proposed a series of solutions to improve the performance of the mining process [4], [5], [7], [15], [16], without eliminating the shortcomings of this method of search data. It became necessary to find adequate techniques and algorithmic solutions to minimize the cost for space and computing time. The Apriori algorithm introduced an approach called "test-and-generate" with pruning. However, this approach suffers from a number of candidates that generates, particularly for relatively small values of support. However, these approaches do not limit the shortcomings of these methods. Given this situation, it became necessary to invest in new methods to faces the following challenges: − Find heuristics to prune the search space; − Find technical or algorithmic solutions, specifically adequate data structures, to minimize the cost in space or in process time.
- 2. 106 Computer Science & Information Technology (CS & IT) We will expose in our present article, the second part of our study (see 3.2 Step 4) and its experimentation that was performed with the basic Apriori in order to demonstrate relevance and efficiency of the approach that we have considered. Later (continuation of our study) we will present the first part (see 3.2, Steps 2 and 3), we adopt Apriori-Cell. 2. RELATED WORK Recent studies have proposed a series of solutions to improve the performance of extraction process of frequent item sets, including cellular automata [10]. Solutions were oriented essentially on the Reduction of I / O and the minimization of the cost of the step of computing the support [9]. Other studies have been based on the discovery of "closed" item sets arising from the theory of formal concepts [11]. Others propose to generate a representative base or generic association rules [14] [16] and used techniques to reduce the number of rules with the use quality measurements [15], syntactic filtering by constraints [2]. 3. THE PROPOSED APPROACH We propose a new approach being located at the junction of two domains that are the Knowledge Discovery from Databases (KDD) on one hand and representation of knowledge from the other. Our approach proceeds in three steps: 1. Extraction of frequent patterns and generating association rules using the algorithm Apriori-Cell which operates on a cell basis; 2. Boolean modelling for association rules; 3. Rules management by the inference engine ICR of the cellular automaton. A cellular automaton is a grid composed by cells which change states in discreet steps. After each step, the status of each cell is modified as states of its neighbors in the previous step [1]. Our approach is implemented by two modules: • The module MAR (Mining Association Rules) ; • The module ICR (Induction of Cellular Rules). The dynamics of the cellular automaton. The inference engine of the cellular automaton simulates the basic operating principle of a classical inference engine using two finite layers of finite automata. The first layer CEL Fact/CEL Item for the basic facts/Items and the second layer, CEL Rule/CEL Transaction for the basic Rules/Transactions. Each cell at time t+1 depends only on the status of its neighbors and his own at time t.
- 3. Computer Science & Information Technology (CS & IT) 107 Figure 1. System Architecture In each layer, the content of a cell determines if and how it participates in each inference step. At each step, a cell may be active (1) or passive (0) i.e whether or not participating in the inference. The principle adopted is simple: − Any cell i of the first CEL Fact/CEL Item layer is considered an established fact if its value is 1, otherwise it is considered as fact to establish. It is presented in three states: input state (EF/EI), internal state (IF/II) and output state (SF/SI); − Any cell j of the second layer CEL Rule/CEL Transaction layer is considered a Rule/Transaction candidate if its value is 1, otherwise it is considered as a Rule/ Transaction which shall not participate in the inference. It is presented in three states: input state (ER/ET), internal state (IR /IT) and output state (SR/ST). − Incidence matrix RE and RS represents the input / output relation of Facts/Items and are used in forward chaining and backward chaining by reversing their order. Thus, the cellular automaton will help optimize the representation of extracted knowledge (association rules) by the boolean principle and their management by using its inference engine through the basic functions δfact and δrule which provide the dynamics of cellular automaton (See 4.1). 3.1.The Proposed Algorithm The process adopted by our system is a succession of four major steps: Step 1: Selection and data preprocessing; Step 2: Cellular representation of preprocessed data; Step 3: Data mining by the cellular automaton using the algorithm Apriori-Cell; Step 4: Post-processing of results. MAR ICR Inductionof CellularRules Induction Graph CELFactsFacts Rules CELRules Cellular Rules Data Source Miningof AssociationRules Association Rules CELTransaction Output Matrix CELItemsetItemset Transaction Input matrix
- 4. 108 Computer Science & Information Technology (CS & IT) Algorithm : Apriori-Cell Input : Transactional-data-base (D), minimum-support S, confidence C Output : lists-of-frequent-items (Fn), Association-rules-base (Br) Fn ← {} Begin Input-matrix = data preprocessing for (D) While we can make joints Do For each line of input-matrix Do calculate support for (item) If support(item) >= S Then Fn← Fn + {item} EndIf EndFor EndWhile Br = Generate-rules (Fn) End 3.2.Principle of the algorithm Apriori-Cell This module applies the cellular principle on the basic Apriori algorithm for mining frequent itemsets. It simulates the basic operating of a join engine inspired by Apriori adapted to cellular automaton, on two finite layers of a finite automaton. the first layer CELItems for the items base, and the second layer CELTransactions for the transactions base. The state of each cell at time t +1 depends only on the state of its neighbors and his own at time t. In each layer, the contents of a cell determines whether and how it participates in each inference step: a cell can be active (1) or passive (0), i.e whether or not participating in the inference. The principle is simple, we suppose that there are l cells in the layer CELItems, and r cells in the CELTransactions layer. The states of the cells are: EI, II, and SI, respectively ET, IT and ST are the input, the internal state and the output of a cell of CELItems, and respectively a cell of CELTransactions. The internal state II of a cell of CELItems indicates the status of the item: in the case of an item, II = 1 corresponds to a state type support_item ≥ minsupport_fixed. For a cell from CELTransactions, the internal state IT will always be equal to 1 (the transactions are always established). The join applied by Apriori-Cell. In the first step the join is made between items using the logical AND, line by line, i.e, it fixes line 1 for example, and it'll do its join to the rest of lines. Once completed, it will go to the second line without considering line 1, this time. And this process continues until the join between items become impossible. At the first iteration, the join is made unconditionally, but beyond 2 items, it applies the following rule: for the join of k-items we must have (k-1)-items in antecedent of the rule to be common. Generation rules. The module Generate-rules is used for the generation and validation of association rules from the lists of frequent n-itemset extracted by Apriori-Cell. This module allows to minimize the number of reading of the database by using to calculate the confidence of each rule, the data cubes, which help in the positioning of the lists of n-itemsets extracted on three dimensions. A dimension for transactions, another one for the antecedent and the last for the consequent of each rule.
- 5. Computer Science & Information Technology (CS & IT) 109 Figure 2. Representation of the data cube, with : x: axis of transactions (1,2) , y : axis of the list 1-frequent items (F1) (aac, acpP, aacA1) and λ : axis of the list l-frequent items (Fl ) Step 4 : post-processing of results (a) Production of induction graph. An algorithm uses as input the association rules {Ri}, items of Antecedenti and Consequenti , and it will give on output an induction graph, with a summit Sp and a node p on which we make a test with possible results binary or with multiple values. (b) Generation of boolean rules from the induction graph. Induction graph is read to generate boolean rules in the following form: Rbk : { Pk } Then { Ck , Sp} (c) Representation of boolean rules. The generated rules (see step 4.b) are represented by cell layers where: { Rbk } gives the set of rules {Rules} and { Pk , Ck , Sp } gives the set of facts {Facts} (d) Integration. The cellular automaton integrates the generated rules in the knowledge base for use through different inference strategies. 4. ILLUSTRATIVE EXAMPLE OF THE REPRESENTATION OF RULES BY CELLULAR AUTOMATA We suppose have obtained the following two rules of association with genes aceA-2, pstS-3, argC and phhB, using the Apriori-Cell algorithm: R1 : {aceA-2=1} , {pstS-3=1} , 45, 77 R2 : {aceA-2=1, phhB=1} , { argC=1 } , 45, 70 and that these two rules have generated the following boolean rules from the induction graph : Rb1: Si { S0 } Alors { pstS-3=1, S1 } Rb2: Si { S1 } Alors { argC=1, S2 } These rules will be represented (step 4, b) in layers CELFacts, CELRules, input matrix (RE) and output matrix (RS). Step 3 : Data mining by the cellular automaton
- 6. 110 Computer Science & Information Technology (CS & IT) From the sample test data base (see 5. Table1 ), we proceed to data mining by Apriori-Cell. We suppose have obtained two association rules with the following genes: aceA-2, pstS-3, argC and phhB. Step 4 : post-processing of results a) Production of the graph induction b) Generation of boolean rules from the graph induction Rb1: If { S0 } then { pstS-3=1, S1 } Rb2: If { S1 } then { argC=1, S2 } c) Representation of boolean rules The boolean rules Rb1 and Rb2 produced are represented by the layers CELFacts (Facts + CELFacts) and CELRules (Rules + CELRules) and input matrix (RE) and output matrix (RS). Facts s0 pstS-3=1 s1 argC=1 s2 CELFacts EF IF SF 1 0 0 0 1 0 0 0 0 0 1 0 0 0 0 Rules Rb1 Rb2 CELRules ER IR SR 0 1 1 0 1 1 RE Rb1 Rb2 s0 1 1 pstS-3=1 0 0 s1 0 0 argC=1 0 0 s2 0 0 RS Rb1 Rb2 s0 0 pstS-3=1 1 0 s1 1 0 argC=1 0 1 s2 0 1 Figure 3. Cell layers of the cellular automaton with input state (EF), internal state (IF) and output state (SF), input state (ER), internal state (IR) and output state (SR) 4.1. Simulation of The Cellular Inference Engine The cellular automaton simulates the operating of an inference engine by using the transition functions previously mentioned, δfact and δrule. Rule antecedent consequent Support % Confidence % R1 aceA-2=1 pstS-3=1 45 77 R2 aceA-2=1, phhB=1 argC=1 45 70 S1 aceA-2 phhB pstS-3=1 S0 argC=1 S2
- 7. Computer Science & Information Technology (CS & IT) 111 We show that simulation starting from CELFacts and CELRules of the example shown previously (Figure 3) by considering that G0 is the initial configuration of the cellular automaton, and ∆ = δrule ο δfact the global transition function: ∆ (G0) = G1. The cellular automaton change state from to the state with if and After application of the law of global transition ∆ we obtain the configurations G1, G2 and finally G3. 1. G0 is the initial configuration of the cellular automaton Facts s0 pstS-3=1 s1 argC=1 s2 CELFacts EF IF SF 1 0 0 0 1 0 0 0 0 0 1 0 0 0 0 Rules Rb1 Rb2 CELRules ER IR SR 0 1 1 0 1 1 2. Evaluation, selection and filtering (application of δfact) Facts s0 pstS-3=1 s1 argC=1 s2 CELFacts EF IF SF 1 0 1 0 1 0 0 0 0 0 1 0 0 0 0 Rules Rb1 Rb2 CELRules ER IR SR 1 1 0 0 1 1 3. Execution (application of δrule) ∆(G0) = G1 Facts s0 pstS-3=1 s1 argC=1 s2 CELFacts EF IF SF 1 0 1 1 1 0 1 0 0 0 1 0 0 0 0 Rules Rb1 Rb2 CELRules ER IR SR 1 1 0 0 1 1 4. Application of the global transition function : ∆ = δrule ο δfact ∆(G1) = G2 Facts s0 pstS-3=1 s1 argC=1 s2 CELFacts EF IF SF 1 0 1 1 1 1 1 0 1 0 1 0 0 0 0 Rules Rb1 Rb2 CELRules ER IR SR 1 1 0 0 1 1 δfact G0 G’0 δrule G’0 G1 G0 G1 G2
- 8. 112 Computer Science & Information Technology (CS & IT) 5. Evaluation, selection and filtering (application of δfact) Facts s0 pstS-3=1 s1 argC=1 s2 CELFacts EF IF SF 1 0 1 1 1 1 1 0 1 0 1 0 0 0 0 Rules Rb1 Rb2 CELRules ER IR SR 1 1 0 1 1 0 6. Application of the global transition function Facts s0 pstS-3=1 s1 argC=1 s2 CELFacts EF IF SF 1 0 1 1 1 1 1 0 1 1 1 1 1 0 1 Rules Rb1 Rb2 CELRules ER IR SR 1 1 0 1 1 0 Final configuration G3 obtained after four iterations. 5. EXPERIMENTATION To examine the effectiveness in practice of our system, we have implemented the engine, and we conducted experimental tests on a machine (Intel Celeron 540 CPU frequency 186 GHz, 512 MB RAM) with a sample test data base (Table 1) representing the genomic sequences mycobacterium tuberculosis) with the first 12 genes of each strain, and the assumption that these genes are sufficiently distinctive and representative of each strain taken separately. Table 1. Test Data Base 1 (12 genes of each strain) 5.2.Discuss of The Results Processing time. We observe that the Apriori algorithm takes an important part in execution time of the system in whole, ie in its most important phases as the generation of association rules by Apriori and the generation of boolean rules. Strain Genes Mt CDC155 aac accD aceA-1 aceA-2 aceB aceE ackA acnA acp-1 acp-2 acpP acpS Mt F11 aceE acpP acpS adk alaS alr argC argD argJ argS aroB aroE Mt H37Ra aac aao accA1 accA2 accA3 accD1 ccD2 accD3 accD4 accD5 accD6 aceAa Mt H37Rv 35kd_a aac aao accA1 accA2 accA3 accD1 accD2 accD3 accD4 accD5 accD6 G3
- 9. Computer Science & Information Technology (CS & IT) 113 Table 2. Evolution of execution time (Basic Apriori and global) Storage space. We find that cell representation is more interesting, and it will be much most prominently with a more substantial sample. Figure 4 . Evolution of storage space 6. CONCLUSION After describing the disadvantages of rule-based methods in data mining, we have proposed an extraction rules guided by a Boolean modeling based on the Boolean principle of cellular automata in order to have a base rules optimized and reduced processing time enough, and thus make a contribution to the construction of knowledge-based systems by adopting a new cellular technic Thus, the advantages of our method based on the cellular automaton can be summarized as follows: • Simple and minimal preprocessing of association rules base, for its transformation into binary matrix according to the principle of cell layers, • Ease of implementation functions δfact and δrule that are low complexity and well adapted to situations with many attributes of rules. REFERENCES [1] Atmani, B., Beldjilali, B. (2007) “Knowledge Discovery in Database : induction graph and cellular automaton”, Computing and Informatics Journal, Vol. 26 N°2 171-197 [2] Besson, J., Robardet, c., Boulicaut, J.F. (2004) “Constraint-based mining of formal concepts in transactional data”, Conference on Knowledge Discovery and Data Mining (PAKDD’04) volume 3056 of LNCS, Sydney- Australia, pp. 615–624 [3] Pasquier, N., Bastide, Y., Taouil, R., Lakhal, L., Stumme, G. (2000) “Mining minimal non-redundant association rules using frequent closed itemsets”, Proceedings of the Intl. Conference DOOD’2000, LNCS, Springer-verlag, July, pp. 972-986 [4] Hajek, P., Havel, I., Chytil, M. (1966) “The GUHA method of automatic hypotheses determination”, Computing 1, pp. 293-308 Confidence % Support % Number of Genes Generated items Number of rules Execution of Apriori Global Execution 10 30 12 37 69 0.00 s 0.00 s 50 50 12 37 125874 0.67 s 1.69 s 70 60 12 37 786756 3.56 s 6.17 s
- 10. 114 Computer Science & Information Technology (CS & IT) [5] Boullé, M. (2007) “Recherche d’une représentation des données efficace pour la fouille des grandes bases de données”, Thèse de Doctorat, Ecole Nationale Supérieure des Télécommunications Paris. [6] Abdelouhab, F., Atmani, B. (2008) “Intégration automatique des données semi-structurées dans un entrepôt cellulaire”, Troisième atelier sur les systèmes décisionnels, Mohammadia– Maroc, 10-11 octobre, pp. 109-120 [7] Agrawal, R., Imielinski, T., Swami, A. (1993) “Mining associations between sets of items in large databases”, Proc. of the ACM SIGMOD Conf., Washington DC, USA [8] Agrawal, R., Srikant, R. (1994) “Fast Algorithms for Mining Association Rules”, Proceedings of the 20th International Conference on Very Large Data Bases, Santiago, Chile, pp 487-499 [9] Bykowski, A., Rigotti, C. (2001) “A condensed representation to find frequent patterns”, Proceedings of the Twentieth ACM SIGACTSIGMOD SIGART Symposium on Principles of Database Systems, ACM Press, pp. 267-273 [10]Fawcett, T. (2008) “ Data mining with cellular automata”, ACM SIGKDD Explorations Newsletter,v.10 n.1, June 2008. [11]Ganter, B., Wille, R. (2004) “Conceptual graphs and formal concept analysis”, Lecture Notes in Computer Science Volume 1257, 1997, pp 290-303Springer-Verlag [12]Mansoul, A., Atmani, B. (2009) “Fouille de données biologiques : vers une représentation booléenne des règles d’association”, CEUR-WS:04-Dec-2009/Vol-547, Conférence Internationale sur l’Informatique et ses Applications CIIA’09, Saida – Algérie [13]Mansoul, A., Atmani B. (2010) “Vers un automate cellulaire pour la fouille de données : Partie I : la représentation booléenne des règles d’association”, ASD 5/6 Novembre 2010, Sfax, Tunisie. [14]Pasquier, N., Bastide Y., N., Lakhal, L. (2000) “Mining minimal non-redundant association rules using frequent closed itemsets”, Computational logic. International conference No1, London, royaume uni (24/07/2000), vol. 1861, pp. 972-986. [15]Vaillant, B., Meyer, P., Prudhomme, E., Lallich, S., Lenca, P., Bigaret, S. (2005) “Mesurer l'intérêt des règles d'association”, Atelier Qualité des Données et des Connaissances (DQK 05, Actes de EGC), pp. 69-78 [16]Stumme, G., Taouil, R., Bastide, Y., Pasquier, N., Lakhal, L. (2001) “Intelligent structuring and reducing of association rules with formal concept analysis”, Proc. KI’2001. conference, LNAI 2174, Springer. AUTHORS Abdelhak Mansoul is an Assistant Professor at Skikda University and affiliated researcher in Oran Computer Lab of Oran University. His research interests are in Database Management System, Data Mining, decision support systems, and simulation. Baghdad Atmani is a professor in computer science at the University of Oran (Algeria). His interest field is Data Mining and Machine Learning Tools. His research is based on Knowledge Representation, Knowledge-based Systems and CBR, Data and Information Integration and Modeling, Data Mining Algorithms, Expert Systems and Decision Support Systems.His research are guided and evaluated through various applications in the field of control systems, scheduling, production, maintenance, information retrieval, simulation, data integration and spatial data mining.