Taxi Recommender System Based On Big Data And Data Mining

The taxi is a kind of transportation that widely used in the modern society and plays an important role in daily life.However,it is difficult to take a taxi sometimes.People don't know where to take a taxi and how long it takes.Nowadays vehicle GPS devices have been installed in many big cities' taxis.The devices can collect the position of the taxis in accordance with a specific rate and send the data to the server system.With the development of distributed computing,it is possible to extract useful knowledge from the big data.The goal of the article is to implement an effective taxi recommender sys-tem based on large-scale taxi trajectory data.Users can get the service by mobile soft-ware with minimal inconvenience.Firstly,this paper introduced the research background and current status.Then,it de-scribed the database,distributed computing and related technology to build the system.On the basis,We analyzed the users' requirement according to the practical application scene and designed the recommender system.The bottom level of the system was con-sisted of the Hadoop distributed system and other databases.In addition,MySQL and MongoDB database stored the data mining result caches.Tomcat was used as a web server for the recommender service.It also supported the carpooling service for multiple user.The paper focused on the data mining module of the system.Then it analyzed the problems of related algorithm and improved them.To solve the problem of miscalcula-tion,this paper purposed a new transfer probability function.It also reduced the pro-gram execution time with the technology of Mapreduce parallel computing framework and the geospatial indexes of quad tree.Experiments had shown that the new algorithm got a better result of matching road and took less time.Then this paper purposed an re-commender model which improved the original algorithm based on probability.In fact,the passengers needed to queue up in the taxi station.Considering the problem,this pa-per introduced the first-in-first-out queue model to describe the scenario.At last,this article gave the experiments on the recommended module and the carpooling module.It was shown that the system got a accurate result compared to the real manual record of vacant cars in different places and the carpooling module also provided an effective ser-vice to save money.The system solved the difficult problem of taking a taxi in daily life so that it was significant and economic.