Personalized Route Recommendation

Recent years have witnessed the rapid development of diverse forms of transportations and Internet,OTA website provide more and more travel information and search service.The increasing number of flights and high-speed trains lead to the information overload problem.Personalized recommendation is a main tool used in solving this problem,and has been intensively applied in industrial practice.Our research is the personalized route recommendation of intercity travel.Route is different from the traditional products which is a combination of different transportations,and dynamically changes in real-time.A route has three characters: fluctuating,diversity and preference sensitive,therefore,departure time,price,airline and other features will directly affect the travelers' buying decisions.Our motivation is to free the traveler from tedious planning tasks,and a personalize route recommendation model which combines different kinds of transportation services is proposed to provide high quality and cost-effective travel route recommendations,taking into consideration individual preferences in real time.Travelers provide the departure time,departure and arrival city,and the model will provide a customized travel route in shorter time.The model includes three parts: search,recommendation and cache.In route search,a directed weighted graph is constructed based on the real-time traffic using the price and time as the weight.With the inputs and requirements of a traveler,an efficient heuristic search algorithm KSPG is designed to generate the top-K candidate routes.In route recommendation,firstly the information entropy is introduced and used to analysis the discrete and continuous features of travelers' history orders.Because of orders' sparse,the traveler similarity function of is defined.With the top-N similar travelers' history order,the entropy is used to calculate the weight of features.Finally the route recommendation algorithm SIERA is designed to rank candidate routes according to the similarity between travelers' preference and routes,and through extensive experimental results show SIERA can effectively improve the degree of satisfaction of the travelers.Route caching can be used to assure the quality of route search and recommendation services.Firstly a city-pairs caching strategy is proposed based on tradition LFU algorithm and time-window.Then for the transit points in city-pairs,the route cache algorithm is designed through sampling and recommendation degree,in route search our model can select the most satisfied transit point with the traveler's preference to reduce the search complexity.Finally through the search time distribution of route search,a route update strategy is proposed to maximize the freshness of cached route ensuring the service quality.