Research On Urban Taxi Service Strategy Based On GPS Trajectory Data

As a beneficial supplement to urban public transport,the number of taxis is increasing year by year,but the gap between taxi drivers' incomes is relatively large.In this case,the research on the relationship between the service behavior pattern of taxi drivers and the driver's income has become one of the hot topics in the field of traffic.The GPS track of taxi reflects the mobile track and traffic condition of taxi drivers and people,which provides a new thought for the analysis of taxi drivers' behavior mode.In this paper,we abstracts the behavior patterns,decision-making process,and hot spot preference in the operation process as service strategy.Based on the GPS trajectory data sets of more than 12000 taxis for one month in Xi'an,this paper studies how to extract taxi service strategy from the trajectory data,and applies it to the establishment of taxi service strategy evaluation matrix.It is also used for the prediction of driver's income and the evaluation matrix of service strategy to analyze the influence of taxi service strategy on taxi drivers' income.The paper proceeds as follows:To begin with,Spark platform was used for large-scale GPS data rankings and trajectory extraction.The realization of data cleaning work,computing of taxi drivers' income and the map grid division of Xi'an city were introduced.Secondly,taxi service strategies and their extraction methods were analyzed and defined,and the service strategy indexes were extracted by different time periods and different grid regions from the GPS trajectory data.In order to find efficient service strategy,the Pearson correlation coefficient analysis method is used to establish the taxi service strategy evaluation matrix,and the Alternate Least Squares(ALS)method is used to the complement of service evaluation matrix.In addition,an evaluation model of taxi service strategy was Constructed based on GPS trajectory data,and the extraction algorithm of taxi service strategy and evaluation matrix was also designed.Besides,the predictability of taxi drivers' income was analyzed based on historical taxi service strategy.Through the Gauss Process Regression(GPR)model,the driver income forecasting model based on taxi service strategy data was designed,and the evaluation matrix prediction model based on the taxi service strategy data was designed by using the deep learning algorithm.Finally,based on the taxi GPS trajectory data,the taxi service strategy and two forecasting models were analyzed in a month of Xi'an with a real and complex situations.The results show that the service strategy has a great influence on taxi drivers' income in different time periods and different regions.The efficient service strategy can be found according to the taxi service strategy evaluation matrix.Based on the taxi service strategy data,the prediction of the driver's income by means of Gauss process regression model have a high accuracy rate,and the 95% confidence interval is given,and the traditional prediction is improved.The prediction of taxi service strategy based on deep learning algorithm works well,and the forecast results can be used to understand the trend of service strategy in the next period of time.