基于ABC优化BP神经网络的船舶交通流量预测

Integration of BP Neuro Network with Artificial Bee Colony algorithm for Traffic Flow Prediction

摘要: 为提高船舶交通流量预测的准确性,针对BP神经网络随机确定初始权值和阈值的缺点,提出一种基于人工蜂群算法(Artificial Bee Colony,ABC)改进BP神经网络的ABC-BP船舶交通流量预测模型。利用人工蜂群算法全局搜索能力和不易陷入局部最优的特点,对BP神经网络的初始权值和阈值进行优化,并以青岛港船舶交通流量统计数据为例,进行实例验证。结果显示,与传统的BP神经网络以及遗传算法(GA)优化的BP神经网络预测模型相比,ABC-BP模型平均绝对百分比误差(MAPE)低至3.361 8%,不仅避免了局部最优,而且通过简单的参数设置就能够显著提高船舶交通流量的预测精度。表明本模型在船舶交通流量预测上是有效可行的。

Abstract: Artificial bee colony algorithm is introduced into the BP neuro network to optimize the initial weights defining and threshold setting. The neuro network for prediction of traffic flow is verified with traffic data from Qingdao Port. The improvement of 3.3618% in mean absolute percentage error is demonstrated compared to BP network only.