PPCI Algorithm for Mining Temporal Association Rules in Large Databases

In this paper an attempt has been made to develop a progressive partitioning and counting inference approach for mining association rules in temporal databases. A temporal database like a sales database is a set of transactions where each transaction T is a set of items in which each item contains an individual exhibition period. The existing models of association rule mining have problems in handling transactions due to a lack of consideration of the exhibition period of each individual item and lack of an equitable support counting basis for each item. As a remedy to this problem we propose an innovative algorithm PPCI that combines progressive partition approach with counting inference method to discover association rules in a temporal database. The basic idea of PPCI is to first segment the database into sub-databases in such a way that items in each sub-database will have either a common starting time or a common ending time. Then for each sub-database, PPCI progressively filters 1-itemset with a cumulative filtering threshold based on vital partitioning characteristics. Algorithm PPCI is also designed to employ a filtering threshold in each partition to prune out those cumulatively infrequent 1-itemsets early and it also uses counting inference approach to minimise as much as possible the number of pattern support counts performed when extracting frequent patterns. Explicitly the execution time of PPCI in order of magnitude is smaller than those required by the schemes which are directly extended from existing methods.