Research On Financial Early Warning Of Listed Companies In China Based On Deep Learning

As typical representatives of the real economy,publicly traded companies,when they are experiencing financial difficulties,not only suffer significant losses themselves,which seriously damages the interests of investors,but they can even have a significant impact on the enconomy as a whole and society in regular development Only scientific research on the identification and early warning of financial difficulties of publicly traded companies will be able to detect the causes of financial difficulties in time and to take effective preventive measures to avoid irreparable losses.The search for an effective and healthy method to prevent financial difficulties is therefore of great importance,both for our economy and for the development of publicly traded companies.This article focuses on standardized research method for research on early impairment of corporate finances and attempts to explore new methods of predicting companies' financial difficulties.After reviewing research already done by national and foreign academics on early financial alerting,it lays out the basis for research on early financial alerting;This article defines the understanding of the financial difficulties of publicly traded companies by analyzing the links between financial risks and the financial difficulties of companies;A financial alert framework has been put in place by analysing internal and external pathways of financial crisis in publicly traded companies;After understanding the design principles of long-and short-memory neural networks(all of which are represented by "LSTM neural networks")adopted in this article,in combination with the TensorFlow framework for in-depth learning and the python-3.6.4 development platform,the programming enabled the processing of relevant data using LSTM neural networks;Analysis of the benefits of applying LSTM neural networks to early financial alert research in our country's publicly traded companies.The internal factors that influence the financial situation of companies were then fully taken into account in the main chapter of this article and a system of financial alert indicators comprising 28 financial indicators and two non-financial indicators has been put in place.A four-tier network model including the Droupout layer and the BN layer(Batch Normalization)was developed using the LSTM neuron network in in-depth learning,while helping companies analyze their financial situation in a way strengthening the selection of early warning variables through the powerful characteristic learning function of in-depthlearning.Then,financial and operational data of listed companies in China during the period 2015-2018 were collected to carry out a two-dimensional financial alert study,i.e.one and several steps,to determine the accuracy of the forecasts of the in-depth learning network model in the different dimensions.The horizontal and vertical comparison of the forecast results shows that the accuracy of the early warning results is all the greater as the period is close to the financial crisis and the cumulative data for the two years preceding the financial crisis are the most accurate for predicting the financial difficulties of listed companies.