Research On Financial Risk Early Warning Of Manufacturing Listed Companies Based On Support Vector Machine

Manufacturing industry is a pillar industry in our national economy and an important force driving scientific and technological progress and social employment.The level and quality of its development has a direct bearing on people’s livelihood.In recent years,Chinese manufacturing industry is faced with the double pressure of fierce international competition and domestic industry transformation and upgrading,which adds more uncertainty factor to the production and operation of manufacturing enterprises,and increases the probability of financial risks.Therefore,it is necessary to establish a set of effective financial risk early warning model,timely and accurately identify risks,to better guarantee the healthy development of manufacturing industry.As far as our country is concerned,after perfecting the securities market access system and withdrawal system,enterprises with continuous losses and serious financial risks will be warned by the CSRC by "ST" or "*ST".Therefore,the construction of financial risk early warning model can not only maintain the healthy development of manufacturing enterprises,but also avoid serious losses to investors and creditors.In this paper,special treatment(ST or *ST)by China Securities Regulatory Commission due to financial problems is taken as the criterion for determining the occurrence of financial risks.The listed manufacturing companies of Shanghai and Shenzhen A-share main board that are specially treated for the first time between 2017 and 2021 are selected as the financial risk samples,and the financial health samples are selected according to the sample pairing principle of the same year,same industry and same region.On the basis of existing research,based on the analysis of the current situation of the manufacturing industry in our country,the selection of 38 financial indexes and 11 non-financial indexes to construct a financial risk early warning index system.The index data of T-3 and T-2 years before the occurrence of financial risks were obtained respectively.K-S test,T-test and M-W U test were used to screen out indicators with significant ability to distinguish positive and negative samples,and feature selection algorithm was used to further obtain the index system most relevant to the empirical purpose.In order to ensure the objectivity of model parameter selection,this paper uses grid search algorithm,particle swarm optimization algorithm and genetic algorithm respectively for parameter optimization and builds a support vector machine model based on it.By comparing the prediction accuracy and robustness of different models,the optimal model is selected.The results show that the support vector machine model(GASVM)with T-2 year data set as input variable and genetic algorithm for parameter optimization has the best prediction performance and higher practical value.On the basis of this model,this paper obtains the region-specific early warning index set from the perspective of regions,constructs the region-specific early warning model based on GASVM,and then puts the sample data sets of each region into the region-specific early warning model and the previously constructed national general early warning model respectively for prediction and model evaluation.The results show that compared with the national general early warning model,the accuracy and robustness of the districtspecific model are higher in the four regions.In general,using support vector machine model to quantify and warn financial risks is of great value for theoretical research and financial risk prevention of manufacturing real economy.