Research on Analyzing Influencing Factors and Predicting Methods of Data Resource Prices Based on Interpretable Machine Learning

Aiming at the current problem of the lack of differentiated and reasonable pricing proposals for data resources in different industries, we propose a data resource price prediction method based on interpretable machine learning by analyzing the influencing factors. Based on the dataset transaction information of China Guoxin Youyi data website, we propose a data resource price prediction and feature analysis model based on XGBoost and SHAP. The best prediction model is obtained by analyzing and preprocessing the raw data; then the processed data are used as inputs to XGBoost and iteratively trained; then the SHAP results are used to analyze the influencing factors on the datasets of different industries, and the combination of the predicted features is optimized using the last recursive deletion method. The results show that the XGBoost-based data resource value prediction model outperforms five major comparison models after tuning the hyperparameters; the introduction of SHAP model enhances the interpretability of the model to identify and characterize the key factors affecting the price of data resources in different industries; the best combination of predictive effect features in each industry is obtained through the last-rank recursive feature deletion; and the best prediction model is developed for the data trading It provides decision-making reference for data trading websites to efficiently filter high-quality and high-value data and increase revenue.