The maximal frequent itemsets mining problem was studied and an algorithm based on pruning itemset lattice effectively was proposed. The itemset lattice tree data structure was adopted to translate maximal frequent itemsets mining into the process of depth-first searching the itemset lattice tree. One of the key measures to promote performance of the algorithm is to prune the itemset lattice tree while traversing it. Three properties of itemset lattice tree were given and three pruning mechanisms, direct superset pruning, indirect superset pruning and transaction sets equivalence pruning, were proposed based on them respectively to prune the infrequent nodes and their extension nodes to reduce the number of nodes while traversing the itemset lattice tree. Test results indicate that all the three pruning mechanisms can reduce the search space effectively and the transaction sets equivalence pruning has the best effect on performance of the algorithm. Test results also indicate that performance of the algorithm is related to denseness of the datasets.