Abstract—Sequential access pattern mining aims to discover interesting and frequent patterns from web data. Most of the sequential pattern mining algorithms are mainly Apriori based and Pattern-growth based. Various algorithms based on Apriori based technique bear the cost of multiple scans of database. Some of the algorithms based on Pattern-growth technique such as PrefixSpan, requires construction of projected databases. WAP-tree based mining techniques require reconstruction of large numbers of intermediate WAP-trees during mining process, which is very costly. In this paper, we propose an efficient sequential access pattern mining algorithm, based on CSB-mine [1]. The proposed algorithm focuses on constructing Web Access Sequence (WAS) list, Unique Symbol (US) list, and generation of SAP table without using WAP trees at any stage. The algorithm eliminates the use of any separate single sequence testing algorithm and it does not need any extra data structure to find first appearance of each symbol, thus saving the space. Also use of compact data structure avoids the reconstruction of projection database, which also saves space and time. The experiments are carried on synthetic data set and we present the performance of proposed algorithm considering memory utilization and run time. Experimental results show that the proposed algorithm outperforms the PrefixSpan and CSB-mine. The results show significant improvement in average memory usage and 10% to 15% improvement in the run time.

Index Term—web usage mining, sequential pattern mining, frequent patterns, prefixspan, CSB-Min

Cite: Krishnakant P. Adhiya and Satish R. Kolhe, "An Efficient and Novel Approach for Sequential Access Pattern Mining," Vol. 7, No. 1, pp. 5-12, November, 2015. doi: 10.12720/jetwi.7.1.5-12