pSPADE : mining sequential pattern using personalized support threshold value

Abstract

As the web log data is considered as complex and temporal, applying Sequential Pattern Mining technique becomes a challenging task. The min sup threshold issue is highlighted - as a pattern is considered as frequent if it meets the specified min sup. If the min sup is high, few patterns are discovered else the mining process will be longer if too many patterns generated using low min sup. The format of web log data that creates consecutive occurring pages has made it difficult to generate frequent sequences. Also, as each user' behaviour is unique; using one min sup value for all users may affect the pattern generation. This research introduced a personalized minimum support threshold for each web users using their Median item access (support) value to curb this problem. The pSPADE performance was the highest on the discovery of user's origin and also interesting pattern discovery attribute.