Research On Mobility Mining And Recommendation Method For Taxi Location Data

With the rapid growth of economy and the continuous improvement of the overall quality of life,people’s travel have a huge growth than before in the number and distance because of work,life and entertainment.Taxis have now become an integral part of the city’s public transport,which has received more and more attention.Today,almost all taxis are equipped with GPS,collecting and using these real data to improve the taxi drivers’ driving strategy is an important and challenging work.For taxi drivers,especially novice drivers,they are not familiar with the overall city’s roads and environment,these will cause low seeking efficiency and increase their fuel consumption,and resulting in an imbalance of income.At the same time,the development of Taxi-hailing apps also makes the competition more intense for taxi drivers,the degree of risk has increased significantly.To solve above problems and improve seeking efficiency,revenue efficiency and income of taxi drivers,we collect a large number of taxi location data and establish data mining and analysis platform on taxis.By mining and analyzing taxi location data,we propose multiple recommendation methods to help the taxi drivers to improve their seeking efficiency,revenue efficiency so as to achieve the purpose of income improvement.The main work of this paper is as follows:(1)Based on the detailed location and direction of taxis,in order to improve the total income of taxi drivers by using seeking efficiency and passenger density of road segment,this paper presents an optimal cruising route algorithm based on the weight of road segment,which is used to find the optimal cruising route for taxi drivers.The accuracy of the proposed method is proved by comparing different recommendation methods under different conditions.(2)To explore how to improve the revenue efficiency of taxi drivers(income of unit operation time),this paper establishes a Markov Decision Process model for passenger search process.For each time period,by learning different MDP parameters from data to find the optimal cruising route for empty taxi drivers,so as to improve their revenue efficiency.A lot of experiments are conducted to verify the effectiveness and scalability of proposed method.(3)We extend the MDP method to the spatial network.Besides,this paper proposes a method of dynamic sliding window to learn parameters from the data and apply it to the model.Taking into account the working day and the rest day,we carry out an extended experiment to verify the effectiveness and accuracy of the proposed approach.A lot of simulation experiments are carried out to analyze the results of each recommendation,these results prove that the proposed approach can effectively improve the operational efficiency of taxi drivers in different situations and reduce the rate of empty driving,consumption of fuel and so as to improve the income of taxi drivers.A lot of experiments are conducted to verify the effectiveness and scalability of proposed methods.