In view of the deficiency of the existing weighted association rules mining algorithms which are not applied to deal with matrix-weighted data, a new pruning strategy of itemsets was given and the evaluation framework of matrix-weighted association patterns, SRCCCI (Support-Relevancy-Correlation Coefficient-Confidence-Interest), was introduced in this paper firstly, and then a novel mining algorithm, MWARM-SRCCCI (Matrix-Weighted Association Rules Mining based on SRCCCI), was proposed, which was used for mining matrix-weighted positive and negative patterns in databases. Using the new pruning technique and the evaluation standard of patterns, the algorithm could overcome the defects of the existing mining techniques, mine valid matrix-weighted positive and negative association rules, avoid the generation of ineffective and uninteresting patterns. Based on Chinese Web test dataset CWT200g (Chinese Web Test collection with 200GB web Pages) for the experimental data, MWARM-SRCCCI could make the biggest decline of its mining time by up to 74.74% compared with the existing no-weighted positive and negative association rules mining algorithms. The theoretical analysis and experimental results show that, the proposed algorithm has better pruning effect, which can reduce the number of candidate itemsets and mining time and improve mining efficiency markedly, and the association patterns of this algorithm can provide reliable query expansion terms for information retrieval.