Online car-hailing demand prediction method based on deep neural network

Abstract

The invention discloses a method for forecasting online car-hailing demand based on a deep neural network, including: S1, dividing the urban road network into regions, and calculating and counting the demand for online car-hailing in each area; S2, designing the network Car-hailing demand forecasting model: Based on deep neural network, learn and train the spatio-temporal characteristics of car-hailing demand, and combine the influence factors of weather factors on the willingness of car-hailing users to predict the regional car-hailing demand in the next time period ; S3. Train the regional car-hailing demand prediction model according to the historical data, and then use the trained model to predict the online car-hailing demand in each region. By integrating external factors and spatio-temporal correlations, this method uses deep neural networks to express the complex nonlinear spatio-temporal correlation characteristics of online car-hailing demand, and can achieve high prediction accuracy.

Description

technical field [0001] The present invention relates to the cross-technical application field of deep learning and car-hailing demand forecasting, in particular to a method for forecasting online car-hailing demand based on a deep neural network. Background technique [0002] With the rapid development of location-based services and mobile Internet technologies, online car-hailing is gradually becoming an important alternative to urban travel. However, there is a mismatch between the demand and supply capacity of online car-hailing in the space-time dimension. For example, there is a phenomenon of "difficulty in getting a taxi" in the city center during rush hour, while there is excess capacity in the suburbs of the city. In addition, external factors such as weather will also affect people's willingness to travel. For example, the demand for car-hailing on rainy days will increase significantly. Therefore, how to use massive cross-domain data to predict car-hailing demand...

AI Technical Summary

Problems solved by technology

The disadvantage of these techniques is that it is difficult to express the complex nonlinear spatio-temporal correlation between car-hailing demand and external factors

With the excellent performance of the deep neural network in learning the complex feature correlation of massive data, people are encouraged to use deep learning methods for demand forecasting, but the current research methods consider relatively single factors, such as: only using convolutional neural networks (CNN ) considers the space factor, and only uses the long short-term memory network (LSTM) to consider the time factor, so the prediction accuracy needs to be improved

Images