High-Efficiency Itemset Mining (HEIM) is a novel problem with applications for analyzing user behavior. HEIM handles the investment of each product, which is significantly effective in the field of retail business. Although the HEIM algorithms can solve problems with high performance, they are not concerned with the hierarchical information (taxonomy) of products in the retail system. Taxonomy is important data and can reveal a lot of helpful knowledge that traditional algorithms have ignored. This paper proposes the problem of Multi-level Efficiency Itemsets Mining (MLHEIM) from a database containing taxonomy data. Furthermore, the paper explains how to calculate the investment of generalized items from their leaf. To solve the MLHEIM, the paper proposes an algorithm called “method for Multi-level High-Efficiency itemset mining - MLHEM”; the algorithm is a combination of search space pruning techniques such as the MLEECS, which is an extension of the EECS. Based on the experimental result, the MLHEM algorithm can effectively execute the MLHEIM problem, and the experimental results will be even better when combined with the proposed strategies.