Research On Value Evaluation Of Data Assets Based On Machine Learning

With the in-depth integration and rapid development of digital information technology,the huge economic and social value contained in data has become increasingly prominent,and it has become a necessary production factor for the development of the digital economy.Using data as an asset is becoming an unstoppable trend,accompanied by the question of how to accurately assess the value of data assets.Given the particularity of data assets,their uncertain and unquantifiable indicators make it more difficult to assess the value of assets.Taking the value evaluation of data assets as an entry point,this article takes data assets as the research object.Through generalizing and summarizing domestic and foreign literature,the concept,characteristics and value of data assets are clarified.Based on these theories,The applicable methods of value evaluation for data assets are analyzed.After combing and comparing traditional asset evaluation and improvement methods,machine learning is selected as the method of data asset value evaluation,and the model is analyzed by using first-hand data from Youedata.First,use the random forest to analyze the main influencing factors that affect the value of data assets,and find that data capacity,data size,data quality,freshness and industry have a decisive effect on the value of data assets.Then establish a multiple linear regression model to predict the value of data assets,and the results show there is a significant‘value gap’between data assets in different industries,and data assets in high value-added industries have a higher value.Finally,the machine learning methods are used to establish the BP neural network model and the random forest model,and the parameters selection of the model are demonstrated in detail.Besides,the Root Mean Square Error(RMSE),Mean Absolute Error(MAE)and R square(R~2)of the empirical analysis results of three models are compared,and it is found that the prediction accuracy of machine learning is much higher than the multiple linear regression model,especially the random forest model,which has lower prediction error and higher goodness of fit,showing stronger predictive ability.Research on the evaluation of the value of data assets is conducive to deepening the reform of factor market allocation,improving the data factor market system,standardizing the currently rapidly developing data transaction market,and providing a useful reference for the healthy development of the big data industry in the future.