Summary: The formal definition of data model for dynamic item weight is given, and a new pruning strategy for weighted itemsets, as well as an evaluation framework, support-confidence-correlation-interest (SCCI), of weighted association patterns is proposed. Based on dynamic item weight and SCCI, an algorithm for the mining of weighted positive and negative association rules is presented. With the characteristics of the dynamic item weighted data being taken into consideration, new pruning methods and evaluation standards are used. Effective weighted frequent itemsets, as well as negative itemsets are mined from the massive weighted database by using the proposed algorithm, and valid weighted positive and negative association rules can be mined by means of simple computation and comparison of itermset weight. The experimental results show that, by using the proposed algorithm, the mining time and the number of candidate itemsets are effectively reduced. Interesting association patterms are obtained, and ineffective patterns are successfully avoided. Compared with the existing mining algorithms, the mining efficiency of this approach is greatly improved, and the problem of the mining of weighted negative patterns is solved based on dynamic item weight.