BP神经网络在我国二手车销量预测中的应用
Application of BP Neural Network in the Prediction of Used-Car Sales in China

随着我国经济的发展，汽车行业得到蓬勃发展。本文基于2001年至2021年我国二手车销量，分析了影响二手车销量的8个因素的相关性，通过灰色关联法表明这些因素之间的相关性高，选择其中关联度高于0.6的7个输入指标，确定二手车每年销售量为输出指标，建立BP神经网络模型。根据模型结果可知，三个数据测试集的确定系数均超过90%，接近1。训练过程的预测值与真实值描点图几乎一致。这表明模型的预测误差小，精度高。测试数据集与真实数据的相对误差也很小，证明基于BP神经网络模型的二手车年销售量预测是合理的。也证实了与传统方法回归分析法相比，机器学习的预测精度更高。
With the development of my country’s economy, the automobile industry has developed vigorously. Based on the sales of used-cars in my country from 2001 to 2021, we analyze the correlation of 8 factors that affect the sales of used-cars in this paper. It is shown that the correlation between these factors is high by the gray correlation method. We select 7 input indicators with a correlation degree higher than 0.6 and estimate the annual sales volume of used-cars as the output indicators to establish a BP neural network model. According to the model results, the determination coefficients of the three data test sets are all over 90%, close to 1. The predicted values of the training process are almost identical to the ground truth plots, which shows that the prediction error of the model is small and the accuracy is high. The relative error between the test data set and the real data is also small, which proves that the prediction of the annual sales volume of used cars based on the BP neural network model is reasonable. It is also confirmed that the prediction accuracy of machine learning is higher than that of the traditional method regression analysis.