Prediction of Web Browsing Behavior based on Sequential Data Mining

Abstract

Discovering time-related transaction behavior or patterns is helpful for businesses in suggesting appropriate products to their customers. For web systems, it is important to understand customers’ browsing behavior to design or recommend products or services that customers need. This study proposes an approach for predicting web browsing behavior that integrates the concepts of sequential data mining, Borda majority count, bit-string operation, and PrefixSpan algorithm. By incorporating the concept of Borda majority count and sequential data mining, the proposed approach can discover majority-based priorities of items for recommendation and improve prediction accuracy. In addition, the proposed approach employs the concept of bit-string operation and the PrefixSpan algorithm to increase computational efficiency. This research employs the concept of ensemble methods that combine multiple models to derive improved results. Compared to previous methods, the proposed approach can yield higher prediction accuracy. Moreover, the proposed approach can provide flexibility for decision-makers in adjusting a minimum support level and the number of items for recommendation. The proposed approach can also be applied to many fields.