Research On Financial Risk Early Warning Of High-tech Enterprises

High-tech enterprises are the main force for high-quality economic growth.Benefited from relevant national policies,high-tech enterprises have obtained rapid growth.However,high-tech enterprises have the inherent characteristics of tremendous R&D investment compared with ordinary enterprises,the R&D risks is also relatively high.They especially face more external risks.On the one side,the market competition environment faced by high-tech enterprises is more intense.On the other side of the shield,high-tech enterprises are more directly adversely influenced by the international environment such as Sino-US trade friction,which makes high-tech enterprises increase their operating costs and face higher financial risks.Hence,the financial risk early warning of high-tech enterprises is a critical topic which is worth of study.On the base of a large quantity of literature researches,this paper adopted the Borderline-SMOTE method to deal with the problem of data imbalance,and took ST(specially treated)high-tech enterprises and non-ST high-tech enterprises from 2011 to2020 as research samples to construct training set and test set.This paper used the feature replacement importance test to screen the early warning indicators,and established a concise,efficient and accurate early warning indicator system specially designed for high-tech enterprises based on the indicators selected initially.This paper used three commonly used algorithms as predictive models: Support vector machine,decision tree and random forest algorithm.Five-fold cross-validation was used to make sure that the conclusion of the model was reliable.This paper compared the accuracy of the above three models and apply random forest to predict the financial risk of high-tech enterprise W company.The main conclusions which were drawn from the verification are:(1)Random forest model with an accuracy rate of 82.54% had the best prediction performance among the three machine learning models.(2)This paper paid attention to the interpretability of machine learning models which was often ignored.Partial Dependence Function and PDP graphs were used to explain the consequences forecasted by the random forest model.(3)Random forest was used to predict the financial risk of W company,the conclusion was that W company would be dealt with as a ST company in 2021.Some reasonable measures were given correspondingly.This research is beneficial for managers.It can assist them to handle financial risk early warning,make them be aware of risks,adjust the corporate governance structure and rules timely and turn the company’s finances to a good trend.The paper is also of good use for investors to avoid risks.