A Single Database Scan Approach for Mining Temporally Similar Association Patterns
Vangipuram Radhakrishna, Puligadda Veereswara Kumar, V. Janaki · 2016
Temporal pattern mining deals with discovering set of all temporal patterns of user interest from underlying input temporal database. The patterns which are of interest to user are to be discovered which requires scanning database repeatedly. The patterns which are not of interest to the user are called as outlier patterns. In this paper, we discover the temporal patterns which are of interest to the user by estimating support bound sequences of temporal patterns and then using these support sequences to estimate distance bounds with respect reference of interest. The distance measure chosen is Euclidean. Since Euclidean distance does not consist upper bound, we normalize the distance to make it feasible between bounds 0 and 1. The approach yields similar temporal patterns as that of conventional association patterns which requires multiple scans. In the best case, a single scan is sufficient however, in the worst; it is required to obtain true supports for some of the patterns. To estimate the support sequences we use the formal expressions designed to compute support bounds.