Research On Data-Driven Financial Performance Evaluation And Prediction Of Mixed Ownership Reform Of State-Owned Enterprises

The mixed ownership reform has entered a new stage of "mixed ownership" promoting "mechanism reform",facing new risks.At present,a lot of mixed ownership reform practice data has been accumulated.However,there are few studies on the financial performance evaluation and prediction of mixed ownership reform of SOEs from the perspective of datadriven.How to use data mining technology to quantitatively evaluate and predict the financial performance of SOEs for avoiding the risk of mixed ownership reform in advance,which has great reference value for the practice of mixed ownership reform of SOEs.Firstly,we use factor analysis to extract six performance factors,namely cost,solvency,liquidity,profit,development and operation from 12 financial indicators in the samples of manufacturing listed SOEs of mixed ownership reform.Using these six performance factors as clustering indicators,the SOEs samples of mixed ownership reform are clustered through the improved kernel Kmeans++ method.The clustering results show three financial performance levels: the high performance level with outstanding solvency and profitability;the low performance level with high cost,heavy debt burden,and operating losses;the moderate performance level with large room for profit improvement.Then,the financial performance levels of the three types of SOEs were predicted using an improved ensemble model.We selected financial indicator in the first three years and indicators of mixed ownership reform such as the proportion of non-state shares,equity concentration,diversity of shareholder types,the ratio of non-state seats and the size of directors and supervisors as predictor variables,which were filtered by recursive feature elimination.The prediction results of random forest,neural network,support vector machine and Adaboost are weighted according to accuracy rates.They are then used as new predictor variables to train the final prediction model.The final improved ensemble model has a prediction accuracy of 85.9%and a 3.1% increase in prediction performance.Finally,we conducted comparison between the indicators of mixed ownership reform among three financial performance level groups and analyzed the importance of predictor variables.Taking Gansu Electric Transmission & Transformer Corporations as an example,the established prediction model is used to predict the performance level of its various mixed ownership reform schemes.The conclusions of our paper are as follows:(1)High non-stateowned shareholding ratio,equity diversity and low equity concentration correspond to high performance;(2)Shareholder diversity,non-state-owned seat ratio,and the size of directors and supervisors are more important than non-state-owned shareholding ratio in prediction of financial performance.(3)The performance level of SOEs that combines equity reform and corporate governance reform is higher.