Classification based on constrained progressive Sequential Pattern mining: A proposed model

Abstract

Sequential pattern is used to get knowledge of data that have time sequence. Sequential patterns contain knowledge that can be useful for users and can be utilized for classification. Sequential patterns become the input for classification process. Research on classification based on sequence was done using Apriori algorithm. However, Apriori takes a long time to search sequential patterns. Moreover, there were also many short and trivial sequential patterns found. Since they are less meaningful, they need to be eliminated. We propose classification based on modified PISA to overcome the problem. Modified PISA is a constrained progressive sequential pattern mining based on PISA. Modified PISA can search sequential patterns in accordance with multiple constraints. Constraints in sequential pattern mining are aimed to get sequential patterns that satisfy user needs. It also will reduce short and trivial sequential patterns that have less meaning to user. Classification based on modified PISA processes data to get sequential patterns which are used as Classifiable Sequential Patterns, CSP, to classify new data. This proposed model will improve classification speed, scalability and accuracy compared to classification based on sequence that utilizes Apriori algorithm.