基于Stacking多算法融合模型的数据资源定价方法训练流程如下

Research on Data Resource Pricing Method Based on Stacking Multi-Algorithm Fusion Model

[目的/意义]针对现有数据资源定价标准模糊、交易规则缺失、买卖双方信息不对称的问题,提出基于Stacking多算法融合模型的数据资源定价方法。[方法/过程]首先,基于四分位法及Box-Cox方法对样本数据进行预处理,剔除异常数据的同时调整数据分布,并基于Lasso算法以均方根误差最小为目标完成特征筛选;其次,基于Pearson系数选取最优算法组合,构建基于Stacking集成学习融合多种机器学习算法的数据资源价格预测模型;最后,以国信优易数据平台交易数据为样本进行算例分析。[结果/结论]相较于单算法模型及其他集成算法模型,XGBoost、SVR、KNN、MLP融合模型的预测精度及性能均明显提升;实现了数据资源自身和市场的双向反馈平衡,形成闭环的价值链,为数据资源定价提供参考。

Purpose/significance]To address the problems of vague pricing criteria,lack of transaction rules and information asymmetry between buyers and sellers,we propose a data resource pricing method based on stacking multi-algorithm fusion model.[Method/process]Firstly,the sample data are pre-processed based on the quartile method and Box-Cox method,and the data distribution is adjusted while eliminating abnormal data,and the feature screening is completed based on the Lasso algorithm with the goal of minimizing the root mean square error.Finally,we analyze the transaction data of Guoxin Youyi data platform as a sample.[Result/conclusion]The results show that the prediction accuracy and performance of XGBoost,SVR,KNN and MLP fusion model are significantly improved compared with the single-algorithm model and other integrated model algorithms;the two-way feedback balance between data resources themselves and the market is achieved,forming a closed-loop value chain and providing a reference for data resource pricing.