Mining Entropy Optimized Parameter based Precise positioning Episode Rules from Event Sequences

Abstract

Episode Rule Mining is utilized in various fields like telecommunication alarm management, stock market, traffic data, customer transactions etc. Episode rule mining is the extraction of important and useful patterns or episodes from large event databases. The main objective of the existing technique Mining Precise-Positioning Episode Rules (MIPER) is to mine episodes from event sequences. For resolving the enormous energy consumption, the proposed methodology ensures selection of parameter setting as well as a simple and effective technique to correct the fixed-gap. This is done by using bat optimization mechanism and this mechanism generates fixed-gap episodes. The proposed study focuses on two major aspects which are correlation between entropy and optimal generation of parameter selection. These are achieved by the proposed approach Mining Entropy Optimized Parameter-based Precise-Positioning Episode Rules (MEOPPER). In this approach, each event’s occurrence time has been clearly specified. To mine Precise-Positioning Episode Rules (PER), a trie-based data structure is used. The proposed algorithm’s efficiency has been analyzed and studied based on three datasets Retail, Kosarak and MSNBC. The experimental results of the proposed work achieves better performance when compared with the existing algorithms like Mining Precise-Positioning-Enumeration (MIP-ENUM), Mining Precise-Positioning-Trie(Depth First Search) (MIP-TRIE(DFS)) and Mining Precise-Positioning-TRIE(Pruning) (MIP-TRIE(PRU) algorithms.