【摘要】 针对电商销量数据的复杂性和现有预测模型的稳定性及泛化能力不足问题，本研究基于大样本多变量数据，应用随机森林和渐进梯度回归树等机器学习模型进行分析。研究结果表明，相比于广义线性回归、弹性网络、支持向量回归、KNN回归树、决策树、多层感知机、AdaBoost,随机森林和渐进梯度回归树对电商销售数据预测拟合更加精确。相比于广义线性回归、弹性网络等7种传统机器学习算法，随机森林和渐进梯度回归树这两种集成学习的方法对电商销量预测更加精确，且渐进梯度回归树算法拟合效果更好、均方根误差更小，是一种更加有效的电商销量预测方法。更多还原

【Abstract】 In view of the complexity of e-commerce sales data and the inadequacies in stability and generalization abilities of existing prediction models, this study carries out an analysis by machine learning models such as Random Forest and Gradient Boosting Regression Trees(GBRT) based on large-sample multivariate data. The results indicate that Random Forest and GBRT achieve more precise predictions for e-commerce sales data compared to Generalized Linear Regression, Elastic Net, Support Vector Regression, KNN Regression Trees, Decision Trees, Multilayer Perceptron and AdaBoost, and the ensemble learning approaches of Random Forest and GBRT are more precise in predicting e-commerce sales, with GBRT showing superior fitting results and lower root mean square error, being a more effective method for e-commerce sales forecasting.更多还原