Constraint-Based Multidimensional Frequent Sequential Pattern in Web Usage Mining

Sequential Pattern Mining is one of the important approaches, which extracts frequent subsequences as pattern in a Sequence Database. Basic formulation of the frequent sequential pattern discovery problem assumes that the only constraint to be satisfied by discovered patterns is the minimum support threshold. Data mining systems should be able to exploit such constraints to speed-up the mining process. Though much work has been done in this area on one and two-dimensional database, mining sequential patterns from multidimensional database is yet on progress. In this paper we introduce an efficient strategy for discovering Web usage mining is the application of data mining techniques to discover usage patterns from Web data, in order to understand and better serve the needs of Web-based applications. Web usage mining consists of three phases, namely preprocessing, pattern discovery, and pattern analysis. This paper describes each of these phases in detail. 
The main objective of multidimensional sequential pattern mining is to provide the end user with more useful and interesting patterns. To mine such kind of sequence data, we have used an extended version of the prefixspan (EXT-Prefixspan) algorithm to extract the Constraint-based multidimensional frequent sequential patterns in web usage mining. A web access pattern is a sequential pattern that is pursued frequently by users. Using these sequences as prefixes a projected database is constructed which is then recursively mined to find the frequent sequential patterns. The EXT-Prefixspan mines the complete set of patterns but greatly reduces the efforts of candidate subsequence generation. Moreover, prefix –projection substantially reduces the size of projected database and leads to efficient processing. We show that the EXT-Prefixspan algorithm is more flexible at capturing desired knowledge than previous Algorithm.