Abstract: Discovering maximum frequent itemsets is a key issue in data mining applications.This paper proposes an iterative algorithm called DIA（Datasets Iteration Algorithm） to construct maximum frequent itemsets.For each iteration，the dataset is scanned twice to find all candidate itemsets whose support count≥1 and part of the maximum frequent itemsets.After each iteration，a new and reduced dataset is produced as the input of the next iteration； along with the iteration process，the candidate maximum frequent itemset reaches its completeness progressively.This algorithm works without generating K-1 order candidate itemsets or FP trees，so it differs from the classical algorithms completely； and because the size of the iterative datasets decrease rapidly，the time complexity of the algorithm is also reduced effectively.Tests show that it is more favorable with large volume data and small minimal support.