基于旅客异质性画像的公铁联程出行方案推荐方法

Road-rail intermodal travel recommendations based on a passenger heterogeneity profile

公铁联程是我国重要的城际出行交通方式之一,但基于单一因素排序的城际票务出行方案推荐方法无法满足公铁联程旅客的个性化出行需求。为提升出行效率,该文基于旅客历史出行订单数据构建画像数据库,使用TF-IDF(term frequency-inverse document frequency)、K-means算法探究旅客异质性衍生的公铁联程出行需求差异,依据偏好得分、敏感特性设置奖励函数,使用Q-learning强化学习算法构建基于旅客异质性画像的公铁联程出行方案推荐方法。以天津-泗洪作为典型的特大城市-小城市公铁联程出行路线,与传统的城际出行方案推荐方法对比,为3类不同敏感特性的旅客推荐公铁联程出行方案。结果表明:该文推荐的公铁联程出行方案能够缩短20%的行程耗时,降低32%的行程费用,在契合旅客行为偏好和敏感特性、满足个性化出行需求方面均有较好的表现。

Road-rail intermodal travel is one of the important intercity travel modes. However, an intercity travel recommendation method based on single factor ranking cannot satisfy the personalized travel demands of road-rail intermodal passengers. This study improves travel efficiency by using a profile database based on passenger historical ticketing data with the term frequency-inverse document frequency(TP-IDF) and K-means algorithms to explore the road-rail intermodal travel demand differences derived from the passenger heterogeneity. The model uses reward functions based on preference scores and sensitivity characteristics with the Q-learning reinforcement learning algorithm in a road-rail intermodal travel recommendation method based on the passenger heterogeneity profile. The method is applied to the Tianjin-Sihong route as a typical road-rail intermodal travel route from a megacity to small cities with road-rail intermodal travel schemes recommended for three types of passengers with different sensitivities. The results show that the recommended travel schemes shorten travel times by 20% and reduce travel costs by 32% while effectively meeting passenger behavior preferences, sensitivity characteristics and personal demands.