Bike-sharing Demand Prediction based on Artificial Intelligence Algorithm Using Weather Data

To alleviate traffic congestion and environmental pollution in heavily populated Seoul, South Korea, the city has actively introduced a bike-sharing service. As the demand for this service increases, efficient supply and management of shared bikes are becoming critically important. Because bike-sharing demand correlates with the weather, we performed a prediction for bike demand based on artificial intelligence algorithms utilizing weather data. The Multiple Linear Regression, Random Forest Regression, and Multi-Layer Perceptron artificial intelligence algorithms were considered and applied to processed weather data collected in Seoul using various methods. The results revealed that the best performance in predicting bike-sharing demand was achieved by a model trained by the Random Forest Regression algorithm, which considers historical weather and demand data and separates itself into two sub-datasets by weekdays and weekends, but without distinguishing itself with the information of the four seasons.