Route Recommendation Based On Public Transportation Card Big Date

With the fast development of public transportation and wide spread use of public transportation card,card data grow explosively.Card data not only contain passenger payment data,but alse contain massive historical trip information.Using these data can help construct good public transporation operation mode.As an important part of public transportation,route recommendation has significant meaning for citizens.However,traditional route recommendation methods are based on cost function estimation and lack support from practical user behaviors.If we can use real massive historical trip behaviors to find practical routes to recommend,the recommendation result will be more close to the real route passengers take.Our work is based on public transportation card data,but the bus transaction records do not contain boarding and alighting stations,which bring great difficulties on understanding passenger behaviors.Considering the high-volume and low-quality nature of the data,we first proposed a preprocessing algorithm based on Map Reduce distributed framework.Then we proposed an algorithm to infer stations based on local information and global information to solve the data loss problem.Finally we proposed a historical behavior based bus route recommendation algorithm to produce routes.Main contributions of our work can be summarized as follows:· Data preprocessing based on public transportation card data We propose a data preprocessing algorithm based on public transportation card big data.We use Map Reduce distributed framework to deal with data,including segmenting transaction records,filtering noise,and labeling time state for trips,to improve data quality significantly.· Passenger boarding and alighting station inference We proposed a station inference algorithm which integrates local information with global information in card to infer stations.The algorithm considers station proximity in both adjacent rides and historical trips with similar start time,and improves station inference accuracy.· Route recommendation based on historical travel behaviors We proposed a route recommendation algorithm based on public transportation card data.In exsiting algorithms,the recommended routes always lack support from practical user behaviors.The proposed algorithm can solve the problem and let the recommendation result be more close to the real routes passengers choose.