基于XGBoost算法的混合模型在零售业数据中的应用
Application of Hybrid Model Based on XGBoost Algorithm in Retail Data

本文使用德国零售数据，经过观察数据的特征，展开数据预处理，建立特征工程；通过将聚类分析与XGBoost模型相结合，建立混合预测模型对德国零售业数据进行建模预测。该模型首先对特征集降维之后选择最优的聚类数，然后对聚类分析后不同的类别分别进行XGBoost模型训练，最后将通过加权求和得到预测结果。研究结果表明，相比较于其他模型，混合模型提升了预测精度和泛化能力。
The paper uses German retail sales data, after observing the characteristics of the data, carries out data preprocessing, and establishes feature engineering; by combining cluster analysis with XGBoost model, a hybrid prediction model is established to model and forecast German retail sales data. Firstly, the model selects the optimal cluster number after dimensionality reduction of feature set, then performs XGBoost model training for different categories after cluster analysis, and finally obtains prediction results through weighted summation. The studies show that the hybrid model improves the prediction accuracy and generalization ability compared with other models.