Fast Weighted Sequential Pattern Mining

In the real world, ordered sequence data is commonly seen, and sequence analysis plays an important role in a wide range of real applications, such as market basket analysis. The weight concept helps to find more interesting sequences, whereas they may be treated as meaningless patterns in sequential pattern mining. Therefore, how to effectively discover these high weighted sequences from a quantitative sequential database is an urgent task. Based on the remaining weight concept, we propose a novel algorithm called Fast Weighted Sequential Pattern Mining (FWSPM) by utilizing an upper-bound called the remaining sequence maximum weight. Based on this upper-bound, an effective pruning strategy is designed to reduce the search space and save memory cost. Experimental results on both real and synthetic datasets show that the designed FWSPM algorithm is more efficient than the existing algorithms, and also has good scalability on large-scale datasets.