Novel Algorithm for Mining Frequent Patterns of Moving Objects Based on Dictionary Tree Improvement

Abstract

Frequent pattern mining is one of the hotspots in the research of moving object data mining. In recent years, with the rapid development of various wireless communication technologies, such as Bluetooth, Wi-Fi, GPRS, 3G, and so on, more and more mobile devices have been used in various applications. In the face of so many data, the following problem is how to deal with and apply the massive data stored in the database, and the theory and technology of data mining emerge as the times require. Therefore, frequent pattern mining for moving objects is very necessary, meaningful and valuable. This paper draws on the existing algorithms for mining frequent patterns of moving objects, and puts forward some innovations. Aiming at the defect of low time efficiency for common frequent mode Apriori algorithm, an improved algorithm named IAA-DT based on dictionary tree is proposed in this paper. The algorithm first traverses all the trajectories in the database and adds it to the dictionary tree. At the same time, it also needs to maintain the linked list to facilitate the pruning of invalid entries. Because of the high degree of compressibility of the dictionary tree, it can be counted directly on the dictionary tree, which greatly improves the time efficiency. The IAA-DT algorithm and the existing improved Apriori-like algorithm based on SQL are compared and analyzed by using the open real data of taxi trajectories in San Francisco. The experimental results with different minimum support levels demonstrate the effectiveness and efficiency of our IAA-DT method.