Research And Application Of Taxi Operation Analysis And Decision Support Based On Hadoop

As one of the important roles of urban passenger transportation,taxis are an important factor influencing urban road traffic planning and management.At present,many urban traffic management departments have huge amounts of taxi data,but the value of these data has not been fully tapped.In order to make effective use of these massive data,this article takes taxi GPS data in the Nantong city as an example,and proposes a taxi GPS data mining method based on the Hadoop platform in order to help the traffic management department or operating company to improve the scheduling efficiency and make the city traffic smoother.People travel faster.According to a company's actual operation of taxi GPS big data,this paper first uses the Spark parallel computer system of Hadoop platform and the HDFS distributed storage mechanism to establish a data model.Through preprocessing the taxi GPS raw data,including coordinate conversion,area division,data cleaning,etc.,the operation indexes such as the number of taxi operations,passenger carrying capacity,and empty driving rate were analyzed.Secondly,this paper statistically analyzes the distribution of taxis in different regions,and studies and discusses the impact relationship between the capacity characteristics of each region and taxi distribution,which includes inflow,outflow,retention,and no-load attributes.The impact of the demand on the car.In terms of data model training efficiency,data storage,and computing performance enhancement,the big data platform is combined with a deep learning framework,and a multi-threaded parallel processing mechanism is added.Python's advanced features and multi-threaded modules are used to improve the synchronization of multi-processing processes.In order to make the load forecasting model more efficient to fit the real situation,two deep learning algorithms DNN and LSTM are put forward which based on TensorFlow framework.The data is processed into the format types suitable for the training of the two models respectively,and the average of the two models is compared.The fitness index is used to choose the optimal model.Finally,this paper has designed and implemented the system's functions,combined with the big data platform and completed the development of the system using Django framework and Baidu map API,and implemented a taxi operation analysis and decision support system based on Hadoop.The system includes data import,Data visualization,evaluation indicators,data forecasts,upload reports,and system assistance.This paper combines the forecasting model with real taxi GPS data and applies it to the analysis and decision-making of actual taxi operations.The related work can effectively solve the problems of efficient operation of taxi drivers and scheduling of taxi managers.Managers provided reliable data support and achieved the goal of assisting decision-making.