Research On Short-term Bus Passenger Demand Forecasting Under Rainy Weather Conditions

Short-term bus passenger demand forecasting is the basis for real-time bus scheduling optimization. Only by accurately grasping the short-term changes in passenger traffic, the dispatcher can reasonably allocate public resources to ensure the balance between supply and demand and the maximum operational efficiency of enterprises. However, short-term bus passenger flow is vulnerable to random factors; its strong time variability and randomness make it not easy to be predicted. And the current short-term bus passenger demand prediction is only based on historical changes of passenger flow, without taking into account the impact of random factors, such as weather. It makes the prediction accuracy is not high. Moreover, the domestic research on the impact of rain and other adverse weather on the bus passenger flow are almost blank. Therefore, this paper fully digs the bus IC card information and weather data. And it analyzes the influence of rainfall weather conditions on the bus passenger. Short-term bus passenger demand forecasting model is established under rainy weather conditions, which is aimed to improve the accuracy and reliability of short-term bus passenger demand forecasting.Public transport IC card data and rainfall data is preprocessed, and this paper makes full use of the bus IC card information, and analyzes the data of the public transportation IC card. Besides, it collects statistics of passengers of bus lines and analyzes the time variation of bus ridership. The influencing factors of public transit passenger flow are systematically analyzed, and the correlation between the precipitation weather factors and the public transit passenger flow is analyzed when the influence of other factors is eliminated. And the impact of adverse weather on the bus ridership is investigated from different types, different periods, and different lines three aspects, obtaining the law of the influence of rainy weather on bus ridership.Based on the analysis of the impact of rainy weather on the bus ridership, a short-term transit passenger demand forecasting model based on recognition similarity model is proposed. This paper applies SVM-KNN algorithm to identify the similarity model, and forecasts the short-term passenger flow. SVM-KNN algorithm combines the Support Vector Machine(SVM) and K-Nearest Neighbor(KNN) algorithm to compensate for the defects that KNN algorithm cannot quickly identify similar patterns in large amounts of data samples and improve the efficiency of the prediction model. SVM-KNN algorithm is mainly based on existing features, directly search the history states which are the most similar to the current state. It is not affected by the parameters, and it can be adapted to the changing environment, improving the reliability of forecasts.According to the actual traffic data, SVM-KNN prediction algorithm is verified. The prediction results are compared with that based on time series model and neural network model. The time series model without considering the rainfall and time series model considering rainfall factor are both contracted. The impact of rainfall on the passenger forecasting and the performance of the SVM-KNN prediction model are tested by comparing the average absolute error, the mean relative error, mean square error, mean square error and mean square error.