Passenger Hotspot Prediction And Route Recommendation Based On Mobile Trajectory Big Data

With the development of the data technology era,mobile trajectory big data analytics has become a research hotspot in urban computing and smart cities.Passenger hotspot prediction is the key to solving the imbalance between the taxi supply and the commuter demand.In particular,accurate passenger hotspot prediction can provide practical decisions for taxi drivers and traffic managers.Meanwhile,route recommendation is a way to help taxi drivers choose better and more efficient cruising routes to increase their profits,especially the optimal route recommendation that benefits both drivers and passengers and improves the efficiency of the urban transportation system.Based on the Spark distributed computing platform and combined with taxi GPS trajectory data,this paper investigates passenger hotspot prediction and route recommendation algorithms and their applications,aiming at reducing taxi drivers’blind cruising expenses,improving cruising efficiency,and maximizing profit.It is dedicated to reducing the traffic load of the city and enhancing its livability.The main contributions of this paper are summarized as follows:1.Data preprocessing.To address the problems of large-scale data storage and calculation in a single machine environment,this paper adopts a distributed processing framework to realize the distributed storage and parallel computing of taxi trajectory data.Moreover,to filter out redundant data and optimize the accuracy of passenger positioning,data processing,data extracting,data sorting,and grid matching are carried out on taxi trajectory data to achieve accurate positioning of passenger hotspots.Finally,combined with the empty vehicle position and the data of road network nodes,the weighted road network undirected graph is constructed,and the road network adjacency matrix is obtained for passenger path recommendation.2.Passenger hotspot prediction.A distributed Ensemble Empirical Mode Decomposition with Normalization of Spatial Attention mechanism Bi-directional Gated Recurrent Unit Model(EEMDN-SABiGRU)is proposed for accurate prediction of passenger hotspots,and it aims at reducing taxi drivers’blind cruising expenses,improving drivers’cruising efficiency and maximizing profits.Specifically,an ensemble empirical mode decomposition normalization(EEMDN)method is proposed to process the passenger hotspots data in the grid to solve the problems of non-smooth sequences and the degradation of prediction accuracy caused by excessive numerical differences,while dealing with the EMD eigenmodal.Then,a spatial attention mechanism is constructed to capture the characteristics of passenger hotspots in each grid,taking passenger boarding and alighting hotspots as weights,and emphasizing the spatial regularity of passengers in the grid.Next,the bi-directional GRU algorithm is fused to deal with the problem that the GRU model can only get the forward information and ignore the backward information to improve the accuracy of feature extraction.Finally,the accurate prediction of passenger hotspots is achieved based on the EEMDN-SABiGRU model using real-world taxi GPS trajectory data under the Spark distributed framework.The experimental results demonstrate that compared with LSTM,EMD-LSTM,EEMD-LSTM,GRU,EMD-GRU,EEMD-GRU,EMDN-GRU,CNN,and BP,the MAPE,MAE,RMSE,and ME values of EEMDN-SABiGRU model decreased by at least 43.18%,44.91%,43.19%,and 39.33%,respectively.3.Passenger path recommendation.A Gurobi optimization algorithm combined with an Angle and an A*algorithm(A~2-Gurobi)is investigated to recommend the optimal taxi route based on big taxi trajectory data in the complex urban road network.Specifically,a road network node extraction method(RNNE)based on the GPS direction of taxis is put forward to solve the difficulty of extracting road network nodes from the big taxi GPS trajectory data.Then,an angle-based sharp point elimination approach(ASPE)is constructed to optimize the searching capability of the Gurobi algorithm to find the shortest path.Next,a Gurobi optimization algorithm(A-Gurobi)based on the A*algorithm is designed,which uses the heuristic function of the A*algorithm to enhance the ability of fast guidance from the origin to the destination,thereby improving the execution efficiency of the Gurobi algorithm.Finally,the A~2-Gurobi algorithm is successfully applied to the optimal taxi route recommendation with real-world taxi trajectory big data.Compared with Gurobi,Dijkstra,Floyd,A*,Bi-A*-ACO,Bellman-Ford,BFS,Acyclic,and AC on the taxi trajectory datasets composed of10,20,30,40,and 50 nodes,the experimental results indicate that the distance of the A~2-Gurobi algorithm for taxi route recommendation is 19.06%,20.97%,3.03%,15.07%,15.07%,and 3.85%shorter than that of Gurobi,A*,Bi-A*-ACO,BFS,Acyclic,and AC on average,respectively,and our algorithm also achieves the same distance as other algorithms.In terms of execution efficiency,our A~2-Gurobi algorithm outperforms the baselines mentioned above by 38.56%,96.63%,66.66%,66.66%,90.87%,97.60%,97.27%,97.21%,and 99.27%,respectively.