Multilevel Interesting Association Rule Mining Using Soft Computing Techniques Dinesh J. Prajapati Associate Professor, Department of Information Technology, A. D. Patel Institute of Technology (ADIT), Gujarat, India. Email: it.djprajapati@adit.ac.in Abstract: Data warehouse contains large amounts of data from a various sources that may contain some noise while using for decision making. Data mining is extraction of knowledge from large data which may contains some amount of missing data along with inaccurate data and outliers. One of the best ways to detect data errors is by properly utilizing association rules that indicates relationships among attributes. Association rule mining algorithms detects patterns which occur in large dataset. Mining association rules at multiple level of concept hierarchy lead to the detection of more specifc and actual knowledge from the dataset. The present paper uses various soft computing approaches for mining multilevel interesting association rules. In real-world problems, transaction data contains quantitative values. The fuzzy logic is useful for fnding interesting association rules in quantitative transactions. To generate optimized multilevel association rule, optimization techniques such as genetic algorithm, ant colony optimization and particle swarm optimization are used. In this paper, soft computing techniques are reviewed based on approach used, fndings and open issues in order to fnd optimized multilevel interesting association rules. Keywords: Ant colony system, Fuzzy logic, Genetic algorithm, Interestingness measures, Multilevel association rule mining, Particle swarm optimization. I. IntroductIon Data mining is used to extract the useful information as well as and patterns from dataset. One of the techniques used in data mining is called association rule mining. Association rule mining is the process of fnding associations or correlations among a set of items or objects in transaction databases, relational databases, and data warehouses. Association rules are of the form A and B à C, where A, B, and C are items. The rule can be comprehended as “Item A and Item B imply Item C”. An itemset is the collection of such items or objects that are being tracked. For example, bread, butter, and jam could be part of an itemset that is of interest for a grocery- food chain. An event that covers the occurrence of one or more Article can be accessed online at http://www.publishingindia.com items from the given itemset is usually known as a transaction [1]. In the case of the grocery-food chain example this could represent a customer buying a set of grocery items. The portion of the rule to the left of the implication (à) is known as the antecedent (A & B), whereas the right side of the implication is known as the consequent (C). Support is the percentage of transactions contains both the antecedent and consequent (P [A, B, C]). Confdence is the percentage of transactions with the antecedent, that also contain the consequent (P [A, B, C | C]). In other words, support represents the frequency of antecedent and consequent items being together in a dataset of transactions, and confdence measures the strength of a rule [2]. Association rules generated from multiple levels of concept hierarchy are called multiple-level or multilevel association rules. In the concept hierarchy tree, each node represents one item of an itemset and terminal node represents actual items appearing in the transactions set. Classes or concepts formed from lower-level of concept hierarchy are represented by internal nodes. At a single concept level, one might fnd that 70 percent of customers that purchase computer may also purchase printer. But, it would be more informative to know that 50 percent of people buy HP printer if they buy desktop computer. The process of discovering such association rules at multiple levels and cross levels, gives us more useful and deeper information about the data set, in comparison to the single level association concept. The purpose of the work described in this paper is to review the soft computing techniques to fnd interesting association rules at different concept hierarchy using minimum support threshold. The rest of this paper is organized as follows: Section II presents preliminaries for soft computing based multilevel interesting association rules generation. In Section III, Introduction to multiple level association rule mining and different rule mining approaches are explained. Section IV shows the literature survey. Conclusion is drawn in Section V. II. PrelImInarIes This section will briefy review fve aspects of literatures. They include fuzzy logic, genetic algorithm, particle swarm optimization, ant colony optimization system and Journal of Applied Information Science 7 (1), June 2019, 01-10 http://www.publishingindia.com/jais