Research And Implementation Of Taxi Recommendation System Based On Multidimensional Measure

Due to the growing popularity of mobile internet. People begin to solve the traditional difficulties of city through the internet, such as the recently appeared taxi software. Through the taxi software, Driver can know the location of the passengers in time. Avoiding the cruising time. Due to the taxi software rules are still in the exploratory. In the beginning, the one who grab the order first will get the passenger. Now the nearest driver will get the chance to get the passenger. This model also has some shortcomings. First, the real distances sometimes are not correct. Second, The nearest driver may can not to drive the passenger to the destination with least time.According to the problem of taxi software, this paper proposes a multidimensional metrics method of driver recommendation. The main content of the paper are:According to the taxi GPS track information, get the road network information, build roads graph of nodes, get the latitude and longitude of each node. On the basis of the road network, study a variety of the nearest neighbors algorithm, IER algorithm, INE algorithm and TDFP algorithm. Based on the advantages and disadvantages of each algorithm, choose the fit nearest neighbors algorithm for taxi situation, and make some improvements. The real distance is one of the dimension metrics. Another measure is the evaluation of the driver. We get the evaluation through the scores on the driver by passengers. We predict scores of passenger to the driver through collaborative filtering algorithm. Thus recommend the best drivers to passengers, by this way, improving passenger service satisfaction.The paper researches and implements the taxi recommendation system based on multidimensional measure. The mainly contribution of this paper is:(1) The paper achieves the improvement nearest neighbors algorithm that fit with the road network, proposes the multidimensional collaborative filtering algorithms to predict the driver score.(2) Through the experiments, show the algorithm is effective and reasonable. And on this basis, we achieve the client and server systems.