First, this paper introduces relation regulation in data mining, then an efficient algorithm SLIG (single-level large itemsets generation) based on relation theory and "AND" operation on recognizable vectors was proposed. SLIG transforms the production process of frequent itemset to the vector calculation process with relationship matrix but only needs to scan the database once. We optimize the algorithm farther, and acquire favorable results. According to this algorithm, combining CRM-instance in the insurance company, we present a detailed process for the solu.tion and analysis of instance, by that we elicit an important conclusion that can bring competition and profit for the company, and also it is a credible gist for reducing risk, at last we analyze the performance among the algorithm of SLIG, SLIG(optimized SLIG) and Apriori.