Passenger Searching Prediction And Passenger Route Recommendation With Mobile Trajectory Big Data

In the data age,when the global positioning system(GPS)technology has become mature,mobile trajectory big data reflect much significant information such as residents' travel characteristics and city operation conditions.Mobile trajectory data have the characteristics of volume,velocity,variety,value,and veracity.In order to solve the problem of distributed data storage and parallel computing,this paper adopts Spark parallel distributed computing framework,combined with taxi trajectory data and road network data,and studies the relevant algorithms and applications for passenger searching prediction and passenger route recommendation.It is aimed at solving the robustness,accuracy,and effectiveness of passenger prediction and passenger route recommendation,providing theoretical basis and technical support for vehicle scheduling,road network design and urban planning.The main contributions of this paper are summarized as follows:1.Data preprocessing: Firstly,in order to solve the problem of large-scale data storage and calculation in the single computer environment,the parallel distributed processing framework Spark is used for distributed storage and parallel calculation of large-scale mobile trajectory data.Secondly,in order to eliminate invalid data and enhance accuracy of passenger positioning,data cleaning,data extraction and data gridding of large-scale mobile trajectory data are used.Finally,the weighted undirected graph is constructed by combining the data of empty vehicle location,passenger location and urban road network nodes(intersections)data;the graph is constructed to obtain the road network adjacency matrix and realize the recommendation of passenger path.2.Passenger searching prediction: A parallel support vector machine optimization algorithm(GS-SVM)based on grid search in Spark framework is proposed to realize the rapid search of passengers in the complex urban traffic network environment.Firstly,the urban road network is gridded on the Spark parallel distributed computing platform to accurately locate passenger hotspots.Secondly,grid search algorithm(GS)is employed to optimize the radial basis function(RBF)of support vector machine(SVM),and the cross-validation method is utilized to find out the global optimal parameter combination,so as to improve the accuracy of passenger hotspot prediction.Finally,based on the parallel distributed framework of Spark,combined with real-world and large-scale taxi trajectory data,we realized the application of parallel GS-SVM algorithm in passenger hotspot prediction.By analyzing seven groups of data sets and comparing with several state-of-the-art algorithms including autoregressive integrated moving average(ARIMA),support vector regression(SVR),long short-term memory(LSTM),and convolutional neural network(CNN),the results of an empirical study indicate that the MAPE value of our GS-SVM algorithm is lower than that of comparative algorithms at least 78.4%.3.Passenger route recommendation: A bidirectional A star ant colony optimization algorithm(BiA*-ACO)is proposed to recommend the fastest passenger path in complex urban road network,combining with passenger prediction results.Firstly,the cost estimation function of bidirectional A star(BiA*)algorithm is used to optimize the heuristic function of ant colony algorithm(ACO),to enhance the global searching ability of the algorithm.Secondly,the optimal path obtained from each cycle is introduced to optimize the pheromone updating rules of ant colony algorithm,to accelerate the convergence speed of the algorithm.Finally,the BiA*-ACO algorithm is applied to recommend the fastest passenger route successfully.By combining the real large-scale taxi trajectory data and the urban road network data,the experimental results demonstrated that when the data set is small,the BiA*-ACO algorithm is at least 47.05%more efficient than the traditional ant colony algorithm.As data integration grows exponentially,under the same starting point,BiA*-ACO algorithm is at least 49.81%more efficient than Dijkstra algorithm and Bellman-Ford algorithm;in terms of path recommendation,compared with A star algorithm and Acyclic algorithm,the fastest path length recommended by BiA*-ACO algorithm is reduced by at least 73.27 m.