Research On Intelligent Recommendation Technology For Taxi-Crusing Based On Big Traffic Data Analysis

In smart cities,a large amount of data is generated everyday,and extracting valuable information from massive data to improve people's standard of living has become a hot topic.With rapid the development of big data technology,it is possible to perform distributed storage and analysis of massive data.At the same time,taxis equipped with GPS positioning system generate a large number of trajectory data every moment,including the basic information of taxis such as taxis number,latitude,longitude,time,cruising status,etc.It is imperative to analyze these trajectory data to achieve smart travel.In the view of the imbalance phenomenon that taxis spend plenty of time cruising passengers,and passengers also need to spend a lot of time waiting for taxis.This paper starts from the perspective of recommending the best passenger-finding strategy for empty taxis,using the big data processing platform Hadoop and Spark to extract and store the histodcal trajectory of Beijing taxis,and combine the real-time information and historical trajectory data of taxi to propose the recommended passenger hotspots,the taxi drivers are provided with the best cruising strategy.In this paper,HRHT recommendation strategy is proposed to provide taxi drivers with the cruising direction,which is divided into two phases:the passenger volume prediction phase and the passenger hotspot recommending phases.In the process of predicting passenger volume,firstly,it is found that the passenger volume are obviously different in the same time period of different dates,and varies on a seven-day cycle.Therefore,this paper presents the Predict Volume of Passengers based on Historical Hotspot(PVHH)model which obtains the distribution of passengers in different time periods in the city through the statistics of passenger volume.Passenger volume prediction model is further trained by the Improved Decision Tree(IDT)classification algorithm.The IDT algorithm weights the passenger distribution data in the PVHH model by analyzing the influence of passenger volume in different time periods on predicting moment.Adaptive Boosting method is used to train the prediction model.Finally,the Spatio-Temporal Index with similar passenger volume is selected based on the result of passenger traffic prediction.In the process of recommending passenger hotspots,this paper proposes the Hotspot Recommendation based on Historical Trajectory(HRHT)which obtains the cruising efficiency and driving time between each two hotspots through the analysis of cruising event and carrying event.And similar passenger volume is selected according to Spatio-Temporal Index.Then,HRHT strategy is used to analyze the selected results to recommend the best passenger hotspot for the drivers.Finally,the historical data of Beijing taxis is used to verify the proposed prediction algorithm and recommendation strategy.Experimental results show that in comparison with KNN algorithm and SVM algorithm,the IDT algorithm has higher prediction accuracy and prediction efficiency,and is more suitable for actual scenario.In addition,HRHT recommendation strategy can accurately recommending passenger hotspots,which save more cruising time and promote the development of smart cities.