基于PageRank算法的出租车需求预测

Prediction of Taxi Demand Based on PageRank Algorithm

智慧交通是智慧城市的重要应用领域之一,大城市的交通问题已经成为了世界级的挑战,城市中出租车的合理分配对交通管理和公共安全具有重要意义。城市交通系统可以看为时变的复杂网络,研究发现应用PageRank计算这种时变复杂网络,再建立自回归积分滑动平均模型(ARIMA),可以预测出下一时间间隔热点的乘客量。实验发现每个地区的PageRank值与ARIMA预测的乘客搭乘量呈正线性关系,因此可将PageRank值作为预测出租车需求的度量,应用北京市10 000辆出租车的行车轨迹数据验证了上述规律。

Smart transportation is one of the important application areas of smart cities. Traffic problem in big cities has become a world-class challenge. The rational distribution of taxis in a city is important for traffic management and public safety. The urban traffic system can be regarded as a time-varying and complex network. The study found that calculating this time-varying complex network based on PageRank, and then establishing an autoregressive integral moving average model (ARIMA), can predict the amount of passengers in the next time interval hotspot. The experiment found that the PageRank value of each region is positively correlated with the passenger occupancy forecasted by ARIMA. Therefore, the PageRank value can be used as a measure for forecasting taxi demand. We verified the above rules by using the trajectory data of 10,000 taxis in Beijing.