基于改进Logistic-SSA-BP神经网络的地铁短时客流预测研究

Subway Short-Term Passenger Flow Forecast Based on Improved Logistic-SSA-BP Neural Network

地铁客流的变化规律存在着一定周期性和潮汐性,针对地铁客流的预测有助于提高城市轨道系统的运营效率,实现轨道交通智慧化运营。为提高地铁短时客流预测结果的准确度,提出了一种基于Logistic混沌映射麻雀算法(Logistic-SSA)优化BP神经网络的地铁客流短时预测模型。该模型通过Logistic混沌映射初始化麻雀算法种群,再利用改进后的麻雀算法优化BP神经网络,达到提高BP神经网络的全局搜索能力和收敛效率;以深圳地铁西乡站进、出站AFC刷卡数据为例,利用构建的预测模型开展客流预测实验,并通过3种准确性评价指标(MAE、RMSE、MAPE),评价改进前后模型预测的准确性。研究结果表明:改进的Logistic-SSA-BP预测模型平均绝对百分误差分别为14.96%和13.73%;与传统BP预测模型相比,其客流预测结果具有更高的准确性。

The change rules of passenger flow of subway have certain periodic and tidal characteristics.Predicting subway passenger flow can help improve the operational efficiency of urban rail systems and achieve intelligent operation of rail transit.To improve the accuracy of short-term subway passenger flow prediction results,a short-term prediction model of subway passenger flow based on Logistic-SSA optimization BP neural network was proposed.The proposed model initialized the sparrow search algorithm population through a Logistic chaotic map,and then optimized the BP neural network by using the improved sparrow algorithm to improve the global search ability and convergence efficiency of the BP neural network.Taking the inbound and outbound AFC card data of Shenzhen Metro Xixiang Station as an example,a passenger flow prediction experiment was conducted by using the constructed prediction model,and the accuracy of the model prediction before and after the improvement was evaluated through three accuracy evaluation indicators such as MAE,RMSE and MAPE.The experimental results show that MAPE of inbound and outbound are 14.96%and 13.73%respectively for the improved Logistic-SSA-BP prediction model.Compared with the traditional BP prediction model,the passenger flow prediction results of the proposed model have higher accuracy.