Comparative Study on Frequent Pattern Mining Algorithms for Temporal Data Set and its Applications

Mining frequent pattern from temporal data set, which contain time related information associated with each transaction is not effectively done by traditional data mining techniques. Because, they mainly focus only on static data. Some patterns are valid only on some particular time points or intervals, so these techniques are not capable of solving over estimating problem of time periods. Frequent pattern of item set in the temporal database is formed, mainly in transactional database like purchasing an item, occurrence of an event etc. Mining time related data is a challenging issue. Different methods for finding frequent pattern on temporal database are studied in this paper and their characteristics like, memory space utilization, computational time, scanning of database, temporal information considered, etc are compared. The three algorithms are, first is Periodic Frequent Pattern Growth algorithm[3] which is an extension of FP-Growth algorithm, give more consideration for periodicity and extract the periodic frequent patterns from PFP-tree, second is Extended a-priori algorithm[1] which is the extension of A-priori algorithm gives more importance for each and every time point and generate frequent pattern on the basis of Time Cube and Basic time cubes, third is the Cluster Based Bit Vector Mining Algorithm[2] which generates the frequent patterns after compressing the database by converting the items in transactions to bit vectors on the basis of occurrence of each item in the corresponding transaction.