基于改进的PSO-BP神经网络的桂林市物流预测研究

Study on forecasting of logistics demand of Guilin based on the improved PSO-BP neural network

由于BP神经网络有收敛速度慢,容易陷入局部最小的缺陷,因此文章提出了一种改进的粒子群算法来优化BP神经网络的权值与阈值。使得预测模型能够在搜索最优解的过程中能够平衡算法的局部搜索和全局搜索,提高搜索的精度。对初步确定的输入指标和输出指标采用线性回归的方法,来筛选与输出指标具有强相关性的输入指标。通过MATLAB软件进行预测,比较标准的PSO-BP与改进PSO-BP模型,预测结果较好,说明改进的PSO-BP模型是有效的。

Because the BP neural network has a slow convergence speed and easy to fall into local minimum defects.In this article,it puts forward an improved particle swarm optimization to optimize the BP neural network weights and thresholds.It enables prediction model to balance the global search and local search in the process of searching the optimal solution.It also improves the accuracy of the search.To find out the primary input and output indicators,this paperuses the method of linear regression.In this way, it can filter a strong correlation between input indicators and output indicators.Through the MATLAB software,it compares the standard PSO-BP and improved PSO-BP model.The improved PSO-BP model prediction results is good.At the same time,it shows that this modelis effective.