Prediction Of Financial Distress Of A-share Manufacturing Companies Based On Integrated Learning

Nowadays,the global economic and industrial structure is constantly adjusting.The integration and upgrading of manufacturing industry and digital science and technology has led to the transformation of industry.In the background of the release of Made in China 2025 and the escalating trade war between China and the United States,China's manufacturing industry is facing great challenges of industrial upgrading.Among them,manufacturing companies maintain a good and stable financial situation.Avoiding financial difficulties in the Sino-US trade war and the increasingly fierce competition in the manufacturing market is crucial for listed manufacturing companies.Therefore,it is necessary to construct a financial distress early warning model that matches the current form of manufacturing industry and effectively realize the risk management and control of financial crisis.Based on the important position and characteristics of manufacturing industry in China's national economy.The innovative design and construction of a financial distress prediction model for A-share manufacturing listed companies fully demonstrate the positive theoretical and practical significance.The main research object of this paper is listed companies of A-share manufacturing industry in China,and the enterprises labeled "ST" or "ST" are defined as listed companies facing financial difficulties.At the same time,the manufacturing listed companies in 2017 are set as T-period,and the financial and non-financial data of 2014(T-3)are used to fit and predict whether they are facing financial distress in T-period.Pearson correlation coefficient method combined with single variable screening index and random forest method based on multi variable screening index.After screening the index system of ensemble learning models,33 financial indicators and 3 non-financial indicators were selected as the final index system of integrated model fitting.1443 non-ST companies and 361 ST companies were generated by using over-sampled imbalance data processing method of SMOTE.Then,three ensemble learning algorithms based on big data technology was used to construct a financial distress predictionmodel.Combining random forest,gradient lifting tree and decision tree model,a two-level learner is designed,which makes the prediction accuracy of the model reach99.99%.Compared with bagging's random forest model and gradient boosting tree model,stacking model has better prediction effect.The stacking ensemble learning model constructed in this paper has better effect on Financial Distress Prediction model.Therefore,it can provide some basis for timely finding out the financial problems of A-share listed companies.