AN EFFICIENT ALGORITHM FOR MINING WEIGHTED SEQUENTIAL PATTERNS

Abstract

Mining weighted sequential patterns is to find higher-value patterns and these patterns can be applied in more fields, and at the same time it addresses some of the storage and resource limitations in the problem of mining sequential patterns with the low min_sup support. The paper proposes a new approach for mining weighted sequential patterns by combining the actual weight values of items in the sequence database with their support to find higher-value sequential patterns set. Moreover, the proposed algorithm uses a vertical database approach, so the algorithm only needs to scan the database once, thus saving execution time. In addition, to increase computational efficiency, the algorithm applies the prime block encoding approach in the computational steps of the extension pattern process. Experimental results show that the proposed algorithm has more effective execution time.