Decision Analysis Based on SVR and XGBoost Prediction Models
Rui Tao, Ziheng Liu, Yajuan Zhang, Xuran Wang, Jingjing Wang, Linlin Li · 2023
This article presents a prediction model for decision analysis based on SVR (Support Vector Regression) and XGBoost (eXtreme Gradient Boosting). The data of question C of CUMCM in 2023 is used, and the attached data is first preprocessed. The Spearman correlation coefficient model is constructed to analyse the correlation of sales between different categories and single products. Based on machine learning analysis methods, the SVR model is built to predict the decision of each category in the coming week. Through the comparison of different models, the XGBoost model is used to predict and analyse the optimal decision of single product, the results show that the coefficient of determination of both the training set and the test set is greater than 0.96, which has a good fitting effect. The sensitivity analysis of the cost markup rate yields a small change in this important parameter for fluctuations of +5% and -5%, indicating that overall the model is less sensitive and has good robustness.