Coteries轨迹模式挖掘及个性化旅游路线推荐

Mining Coteries Trajectory Patterns for Recommending Personalized Travel Routes

Coterie是一种异步的组模式,要求在不等时间间隔约束下,找出具有相似轨迹行为的组模式.而传统的轨迹组模式挖掘算法往往处理具有固定时间间隔采样约束的GPS数据,因此无法直接用于Coterie模式挖掘.同时,传统组模式挖掘存在语义信息缺失问题,降低了个性化旅游路线推荐的完整度和准确度.为此,提出基于语义的距离敏感推荐策略DRSS(distance-aware recommendation strategy based on semantics)和基于语义的从众性推荐策略CRSS(conformity-aware recommendation strategy based on semantics).此外,随着社交网数据规模的不断增大,传统组模式聚类算法的效率受到极大的挑战,因此,为了高效处理大规模社交网轨迹数据,使用带有优化聚类的MapReduce编程模型来挖掘Coterie组模式.实验结果表明:MapReduce编程模型下带优化聚类和语义信息的Coterie组模式挖掘,在个性化旅游路线推荐上优于传统组模式旅游路线推荐质量,且能够有效处理大规模社交网轨迹数据.

Coterie is an asynchronous group pattern that finds the group patterns with similar trajectory behavior under unequal time interval constraints. The traditional trajectory pattern mining algorithm often deals with GPS data with fixed time interval sampling constraints, which cannot be directly used for coterie pattern mining. At the same time, the traditional group pattern mining has the problem of missing semantic information, and thus reduces the completeness and accuracy of individualized tourist routes. To address the issue, two semantic-based tourism route recommendation strategies, distance-aware recommendation strategy based on semantics （DRSS） and conformity-aware recommendation strategy based on semantics （CRSS）, are proposed in this paper. In addition, with the increasing size of social network data, the efficiency of traditional group model clustering algorithm is of great challenge. Therefore, in order to deal with large-scale social network trajectory data efficiently, MapReduce programming model with optimized clustering is used to mine the coterie group pattern. The experimental results show that the coterie group pattern mining with optimized clustering and semantic information under the MapReduce programming model achieves better recommendation quality than the traditional group pattern travel route in the personalized tourism route recommendation and can effectively handle the large-scale social network trajectory data.