Financial Risk Early Warning Model Of Agricultural And Forestry Listed Enterprises Based On Improved AutoEncoder

In recent years,the state has vigorously developed the agricultural industry and issued various favorable policies to benefit and help agriculture.Agricultural related enterprises have gradually become the main role in the overall social and economic environment.Ensuring the stable and efficient development of agricultural enterprises is an important goal.In order to achieve this goal,it is necessary to reasonably avoid financial risks.This thesis constructs a financial risk early warning model based on improved AE,and selects the financial data of 167 companies in agriculture and forestry related industries in recent seven years;The similarity calculation method is used to screen the indexes,and an index system composed of 52 indexes is established;Entropy method,critic method,independence method and information weight method are used to calculate the weight of various financial indicators respectively,and then the weight is combined with grey correlation analysis method,fuzzy comprehensive evaluation method and topsis method to score the company’s financial risk,and k-means clustering method and SOM clustering method are used to classify the financial risk level,The Auto Encoder combined with convolutional neural network is improved to obtain the financial risk prediction model,and the processed data are put into the built model for training and prediction.The experimental results show that the CRITIC-FUZZY-SOM-AE-Res Net model constructed in this thesis can achieve good early warning effect in the field of agriculture and forestry economy.The highest prediction accuracy is 94% and the average prediction value is 87%,while the prediction accuracy obtained by the traditional AE-SVM and Res Net models is only 77% and 79.21%,Therefore,the model proposed in this thesis can become a great help in the field of financial risk early warning of agricultural and forestry listed companies.