Mining closed sequential patterns - A novel approach

Generation of data with an inherent sequential nature is the order of today´s digital society. This kind of data is composed of discrete events that have either a temporal or spatial ordering and is generally obtained by sectors like telecommunication networks, E-Commerce, Internet servers and gene databases, medical domain to name a few. The ability to explore and exploit the sequential nature of the data for prediction leverages strategic decision making and problem solving. Symbolic sequence data consists of long sequence of ordered events with possible relationships among them. Symbolic sequence mining techniques aim at extracting frequent sequential patterns from huge collections of event sequences based on the user defined minimum support threshold. For a given set of symbols / events due to the possible repetition of events an infinite large number of sequences is possible and hence the task of extracting frequent sequences is complex. Whereas in closed sequential patterns, the set of sequential patterns automatically eliminates a lot of redundancy from the set of all frequent sequences and provides a concise set of patterns maintaining completeness. A novel approach is proposed for extracting closed sequential patterns which can be applied to fields that prefer complete and concise number patterns for analysis to aid the process of effective decision making.