An Approach to Mine Time Interval Based Weighted Sequential Patterns in Sequence Databases

Abstract

Sequence pattern mining is an important data mining task with broad applications. Many sequence mining algorithms have been developed to discover frequent sub-sequences as sequential patterns in a sequence database given the minimum support threshold. One of the drawbacks with the conventional sequential pattern mining is, it considered only the generation order of elements in the sequences in finding sequential patterns.However, in real world application domain sequences, the generation times and time-intervals between the elements are also very important. Another drawback is, all the sequence patterns are treated uniformly while in reality different sequential patterns have different importance. To address the second drawback, weighted sequential pattern mining was proposed, which aims to find more interesting sequential patterns, by considering different significance for data elements in a sequence database. However, weighted sequential pattern mining did not consider time-interval information of the sequences. This paper presents a new approach for mining time-interval based weighted sequential patterns (TIWSP) in a sequence database. In the proposed approach, the weight of each sequence in a sequence database is obtained from the time-intervals of successive elements in the sequence, and then sequential pattern are mined by considering the time interval weight. Experimental results show that TIWSP mining is efficient than PrefixSpan in generating more interesting patterns.