Study Of The Stock Pricing Function Form And Relevant Data Denoising Method

Price as the most important signal between buyers and sellers, is the soulof market, so it is particularly important to build reasonable stock pricing function,numerous scholars are in favor of the research in the form of it and put forwardvarious pricing methods. However, lots of current research results still need to beimproved, such as the single selection of variables, functional forms are mostlylinear, etc. Combined with stock price existing noise problems and based on therelated literatures, this paper mainly studies three aspects which are as following:(1) In this paper, through the analysis of stock pricing mechanism, we argue thecorporate level variables(EPS,NAPS)which are used for the present pricingfunction need to be improved, and put forward the corporate level should becombined with the market level, after considerable analysis, we select the Betavalue and the floated cap as the market level, and carry out the empirical analysis.Our results show that the prediction precision of the model based on consideringthe company and the market level variables has more advantages than onlyconsidering the former one.(2) Compared with various forms of production function, since the stock pricingmechanism is complicated, so the stock price should be a nonlinear function form.Accordingly, we select the Cobb-Douglas production function as pricingfunction. Meanwhile, we introduce the statistical test method of regression model,which will be a theoretical basis for later empirical analysis. In the empirical part,compared with the linear regression pricing function with the nonlinear one, theresults show that the precision of our nonlinear model has obvious improvement.(3) We could not evade the noise problem of the stock-price, which as arepresentative of the financial time sequence, so investigating the denoisingproblem is particularly important. Accordingly, we put forward the denoisingmethod based on wavelet analysis, an empirical research is given to show themodel with the denoising stock-price gets higher prediction precision.