Taxi Passenger Flow Prediction Based on Hotspot Clustering and Neural Networks

With the development of society and urban expansion, the increase in urban population has led to growing transportation pressures. Therefore, predicting passenger flow and consequently planning taxi routes have become urgent problems to solve. In this paper, Chengdu taxi GPS trajectory data is utilized. Convolutional Neural Networks (CNN) are used to extract features from vacant taxi periods and locations. Long Short-Term Memory (LSTM) networks and clustering algorithms are employed to predict and determine passenger hotspot areas based on data-driven approaches. By integrating neural networks with temporal and spatial distribution features of passenger flow, this method effectively reflects the dynamics of taxi passenger flow. Simulated experiments are conducted on Chengdu taxi GPS routes. The experimental results demonstrate that the proposed model outperforms existing methods. The findings indicate that the model effectively uncovers the spatiotemporal correlations in taxi passenger flow data, leading to accurate predictions.