Author : Ms. Pratiksha Prakash Ghule and Prof. DR Swati Bhavsar

Frequent itemset mining is important task in data mining domain. This is applicable in variety of applications such as market-basket analysis, browsing history analysis, transaction record analysis, etc.  Lot of work has been done in the domain of co-occurrence pattern extraction and association rule mining.  The existing system works on static dataset as an input. The proposed system focuses on temporal analysis of data and extracts cooccurrence patterns from dataset based on timestamp information. Each record in a dataset has it is own time information.  Based on the time information data is sliced in 3 dimensional cube. The apriori algorithm is extended to work with time cube information data. Using time interval cubes, the co-occurrence of pattern is analyzed periodically and with certain time interval. For processing multiple time cube simultaneously, a multithreaded environment is proposed to improve system efficiency. To solve the overestimation problem, a density threshold value is checked for each time cube. The performance of system is tested on various dataset and its execution time and memory is compared with the existing approach.

Keywords: Co-occurrence patterns, time cube, Temporal analysis, apriori, multithreaded application, frequent patterns