Mmp-Tree for Sequential Pattern Mining with Multiple Minimum Supports in Progressive Databases

Abstract

Although there may be lot of research work done on sequential pattern mining in static, incremental, progressive databases, the previous work do not fully concentrating on support issues. Most of the previous approaches set a single minimum support threshold for all the items or item sets. But in real world applications different items may have different support threshold to describe whether a given item or item set is a frequent item set. This means each item will contain its own support threshold depends upon various issues like cost of item, environmental factors etc. In this work we proposed a new approach which can be applied on any algorithm independent of that whether the particular algorithm may or may not use the process of generating the candidate sets for identifying the frequent item sets. The proposed algorithm will use the concept of “percentage of participation” instead of occurrence frequency for every possible combination of items or item sets. The concept of percentage of participation will be calculated based on the minimum support threshold for each item set. Our algorithm MS-PISA , which stands for Multiple Support Progressive mIning Of Sequential pAtterns , which discovers sequential patterns in by considering different multiple minimum support threshold values for every possible combinations of item or item sets.