基于多源异构和XGBOOST模型的销量预测

Sales Forecast Based on Multi-Source Heterogeneous and XGBOOST Model

随着中国经济的快速发展,中国汽车市场已经发展成为世界第一大汽车消费市场和全球最大的消费潜力市场3面对规模越来越大的市场,准确预测铕量是汽车产生面临的重要课题。本文在主成分分析的技术之上探讨了多源异质性大数据和XGBOOST算法在汽车销量预测模型中的应用首先选取一些社会经济结构变量、相关政策以及某汽车企业汽车销量为样本数据;其次为了去除多重共线性问题利用主成分分析分析出权重占比累积超过90%的样本特征;再次利用多维度控制变量法确定了XGBOOST糢型的最佳参數组合;最后利用训练好的模型对验证集进行测试,结果表明,多个月的预测销量与实际销量之间的平均准确率为88.52%,平均差异仅为11.48%。这为企业预测汽车销量提供参考.

With the rapid development of the Chinese economy,the Chinese automobile market has developed into the world's largestautomobile consumer market and the world's largest consumer potential market.In the face of an increasingly large market,it is an important issue for automobile production to accurately predict sales.This paper discusses the application of multi-source heterogeneous big data and XGBOOST algorithm in the car sales forecast model based on the technology of principal component analysis.First,select some socio-economic structural variables,relevant policies,and car sales of a car company as sample data;second,use principal component analysis toanalyze the sample characteristics of the weight ratio of more than 90%cumulative;again use the multi-dimensional control variable method to determine the XGBOOST model the best parameter combination;finally,the trained model is used to test the validation set.The results show that the accuracy rate between the predicted sales volume and the actual sales volume is 88.52%,and the difference is only 11.48%.This provides a reference for enterprises to predict car sales.This provides a reference for companies to predict car sales.