Abstract: Most frequent item sets mining is the focus and the difficulty of text association rules mining, andit directly determines the performance of text association rules mining algorithms. Firstly, several most frequentitem sets mining algorithms are analyzd and summarized. And then, traditional inverted list is improved. Based onthe improved list and set theory, a new TOP-N most frequent itemset mining algorithm combined minimum supportthreshold dynamic adjustment strategy is presented. In addition, several propositions and deductions for improvingthe performance of the performance of the provided algorithm are offered. Experimental results show that theprovided algorithm is better than Napriori and IntvMatrix.