Abstract: In order to solve the demerits of slow computational efficiency in the process of mining large data sets with the conventional association rule method and mining a large number of redundant rules, a new data mining algorithm based on association rules and similarity (U-APR) is proposed. Firstly, the algorithm reads the data and constructs the matrix at one time, and uses the characteristics of association rule supporting measurement to add judgment attributes and speed up the end of the iterative process, thereby overcoming the problem of frequently scanning the database in the classical Apriori algorithm. Then, it uses the similarity algorithm to delete redundant association rules. Finally, combined with confidence, support and user goal matching, the mining results are sorted and output, so as to obtain the association rules that users are interested in. At the same time, the algorithm and two common association rule methods are used to mine the financial data of students in a university in Guangdong. The experimental results show that compared with the two association rule methods, the U-APR algorithm shortens the operation time and improves the utilization of storage space, exhibiting an optimization effect on the analysis and mining results of users.