Trend modeling and multi-step taxi demand prediction

At present, there is a serious mismatch between the supply and demand of taxis, and reasonable demand forecasting can effectively reduce the supply -demand gap, which is an important foundation for taxi scheduling. This article proposes three modeling methods for taxi demand cycle trends, namely the Fourier series based method, the principal component analysis trend based method, and the average trend based method. Finally, based on a weighted combination of three periodic features, a multistep prediction model for taxi demand was established. On actual data, the method proposed in this paper achieved an MAE error of 1.91, indicating that it can effectively predict taxi multi -step demand. Furthermore, after comparison, the method proposed in this paper outperforms other comparative methods in predicting taxi demand.