Research On Valiue Prediction Model Of Internet Platform Enterprises Based On Support Vector Machines

In recent years,network power,national big data,"Internet +" and other strategies of China are constantly pushing forward the historical process of development of informatization.Internet industry of China is booming,and a large number of excellent Internet platform enterprises have emerged.Internet enterprises are playing an important role in the financial market.In March 2019,data released by the China Academy of Communications showed that the total market value of Internet companies exceeded eight trillion yuan.Although the market value of Internet platform enterprises is very high,in the face of the changeable Internet industry environment,many factors need to be taken into account when predicting the value of platform enterprises.A slight mistake will lead to a complete failure,which is crucial for investors,the public,other enterprises and the development and progress of platform enterprises.However,the traditional income present value method,market comparison method and cost method and other mainstream methods are not fully applicable.Not only need to consider the financial situation of the company,but also pay attention to the unique "flow" data of the Internet platform company.Therefore,determining the value prediction method of these companies is an urgent problem to be solved now.In order to realize the value prediction of Internet platform companies,first of all,this paper uses literature analysis methods to analyze the factors that affect the value of Internet platform companies,and screens out the profitability,operating capacity,solvency and shareholder income,sustainable development capacity 38 indicators in seven dimensions of user capabilities and platform management capabilities.Then studies for 49 Internet platform companies listed in China’s A-share,H-share,New York,and Nasdaq stock exchanges as research objects.using the principal component analysis(PCA)method tofilter the financial and non-financial indicators,draw a set of value system suitable for Internet platform companies.Secondly,interdisciplinary research methods are used to innovatively combine machine learning theory and value prediction,use support vector machines to train the basic model,and optimize and improve the basic model through principal component analysis and grid search algorithms to obtain different combinations.Then get the four models of SVR,PCA-SVR,GS-SVR and PCA-GS-SVR with different methods of combination.The comparative analysis of the prediction results shows that the average prediction accuracy of the PCA-GS-SVR model with multiple optimization methods and the error of a single sample are significantly better than the SVR model and a single method that only considers the grid search algorithm and the principal component analysis method.The combined model achieves the ideal value prediction effect,and also verifies that the machine learning method can be well applied to the prediction of the market value of Internet companies.It also shows that the model has certain applicability for promotion and is suitable for investors and the public.