Research On Prediction Model Of Bus Arrival Time Based On Machine Learning

With the continuous acceleration of urbanization,the contradiction between traffic supply and demand on the road network has also become prominent.The traffic supply is now far from meeting people’s travel needs,urban roads are becoming more and more crowded,and environmental pollution is becoming more and more serious.Among various methods for solving urban traffic problems,vigorously developing public transport is a major means.Accurate and reliable bus arrival time prediction is an important way to improve the service level of the bus system and attract passengers to choose bus travel.Therefore,this paper conducts an in-depth study on the bus arrival time prediction model,divides the arrival time prediction into two parts: the stop time prediction and the travel time prediction between stations,respectively builds the corresponding prediction models and conducts experiments to verify the effectiveness of the models.The specific research contents are as follows:First,this paper analyzes the collected data and preprocesses the data,removes and cleans the problematic data,and repairs the missing data.Further analyze the historical data from the perspectives of all-day running time,the same time period on the same day,and the same time period in history,and study the running rules of buses to provide a basis for subsequent model construction.Secondly,a prediction model of bus stop time and a prediction model of travel time between stations are established respectively.In terms of the station stop time prediction model,a PSO-KNN model is established.After the KNN algorithm is improved by the particle swarm algorithm,the particle position is dynamically optimized in real time,and the calculation efficiency and prediction accuracy are improved.In terms of the travel time prediction model between stations,a GCN-LSTM model for jointly extracting spatiotemporal features is established.After historical data is input,the spatial features are first extracted through the graph convolutional neural network,and then the data with spatial features enters the long-term and short-term memory network.The extraction of time characteristics is carried out in the process,and the prediction of travel time between stations is completed.Finally,the two models proposed in this paper are experimentally verified.The GPS data of K13 bus driving in Liaocheng City,Shandong Province was selected as the experimental data,and the two models were predicted after the data was normalized.After comparative analysis with the actual value,the parking time prediction model and the inter-station travel time prediction model constructed in this paper have a high fit,and the error value is small.By comparing with the existing model algorithms,the accuracy of the model is further tested.After experiments,it is found that the model proposed in this paper has a smaller MAPE value and a higher model prediction accuracy,which is a practical model.