International Journal of Computer Applications (0975 – 8887) Volume 75– No.12, August 2013 1 An Efficient Binary to Decimal Conversion Approach for Discovering Frequent Patterns Kapil Chaturvedi Department of Computer Application Rajiv Gandhi Technical University Bhopal, MP, India Ravindra Patel, Ph.D Associate Professor Rajiv Gandhi Technical University Bhopal, MP, India D.K. Swami, Ph.D Professor VNS Institute of Technology Bhopal, MP, India ABSTRACT Association Rule Mining(ARM) is a most vital field of data mining to discover interesting relationship between items from huge transaction databases it analysis the data and discover strong rules using different measures such as (support, confidence, lift, conviction) etc, various ARM algorithms are available in literature for discovering frequent patterns. Market Basket analysis is one of the most essential applications of ARM; other applications are pattern recognition, weblog data mining and special data analysis etc. In this paper we proposed B2DCARM algorithm to discover frequent pattern which use Boolean matrix based technique. This algorithm adopts binary to decimal conversion approach to discover frequent itemsets from huge transaction database which outperforms in both of the cases where support threshold is low or high and also better performs from efficiency point of view compare to available tree based approaches. Keywords ARM, B2DCARM, Frequent Pattern mining, Boolean matrix 1. INTRODUCTION Data mining is the process of extraction of hidden predictive information from large databases [1, 2], it is also known as knowledge discovery process (KDD) and the major characteristic of data mining is to provide “proactive information delivery from the business perspective”. Data mining is the most popular research field since last two decades, many of mining techniques already exist and well investigated in literature[10], data mining has several applications like DNA pattern recognition, market analysis, web mining etc. Association rule mining is one of the most important and well researched field of data mining, which was first introduced by Zhang C Q, Zhang S C in 1993 [3, 4]. It aims to extract interesting correlations, frequent patterns, associations or casual structures among sets of items in the transaction databases or other data repositories, an association rule can be represent in the form of an implication expression as A  B where A and B are the disjoint itemsets, i.e. A ∩ B= ∅ here support and confidence are two measures for finding the strength of an association rule with additional measurement factors like Lift and Conviction. These terms are briefly discussed below.  Support Determines the occurrence frequency of item/rule in given dataset.  →  =  ∪      Confidence ascertains how frequently RHS of the rule present in the transaction where LHS of rule will also be present (i.e. items in B appear in transactions that contain A).  →  =  ∪    Lift is defines as “ratio of the observed support to that expected (if A & B were independent)”  →  = (∪)  × ()  Conviction is the ratio of the expected frequency of occurrence of A without B, that means “the frequency that the rule makes an incorrect prediction (if A & B were independent)”  →  =  − ()  − (→) For example the rule (PEN, PENCIL)  NOTE BOOK found in the sales transaction database of a stationery shop, would indicate that if a customer buy PEN and PENCIL together would also buy NOTE BOOK such information can be use full in decision making about marketing activities. Traditional ARM techniques can be classified in three categories. 1. Candidate generation approaches [5, 7]. 2. Tree based approaches [6]. 3. Matrix based approaches [8, 9]. There are many algorithms are available and well investigated in the literature based on these techniques. This paper proposes a frame work for binary to decimal conversion Association Rule Mining (B2DCARM) algorithm for discovering the frequent patterns(in section III), section IV shows the experimental result and section V is the conclusion part which shows outcome of experiments performed. 2. RELATED WORK 2.1 Classical ARM approaches Association rules are widely used in various areas such as telecommunication networks, market and risk management, inventory control etc. Typical association rule mining algorithms are Apriori and FP-Growth, first association rule mining algorithm was Apriori algorithm introduced by Agrawal and R. Srikant in (1994) [5] which uses a candidate generation approach to find frequent items. Another