Based On The Value Of The Financial Indicators Of Investment In The Stock Price Prediction Model

Based on the Value Investing theory, this paper considers intrinsic value as the determinant of stock prices. This paper predicts the stock price by nonlinear model and compares it with the linear estimation model by empirical research.First, this paper introduces the status quo of the value investing theory at home and abroad, the assessment model of the intrinsic value of stock, the status quo of China's stock market forecasting. Abroad, Graham pointed out the direction for us in theory, Buffett proved that Value investing can bring excess returns by practice. At home Jianwei Liu, Bing Dan etc. are the researchers and communicators of Value investing theory. Assessment of intrinsic value can be invided into the absolute value Assessment and relative value assessment. This paper focuses on the traditional discounting model. The status quo of China's stock market forecasting shows that a growing number of scholars tend to use intelligence technology used in forecasting the stock market.Then, the neural network is introduced, focusing on the multi-layer feedforward neural network. The neuron activation function is S-differentiable functions which can achieve any nonlinear mapping from input to output. Multilayer feedforward neural networks for nonlinear function approximation are particularly appropriate, so the use of multilayer feedforward neural network to model the dynamics of the stock market is very appropriate. This paper chooses 7 major financial indicators covering the fundamentals to represent the decision factors of the intrinsic value of the stock price, and use the multi-layer feedforward neural network model to predict stock prices.The last part of this paper is the implementation of the prediction model based on the multi-layer feedforward neural network. The implementation method is divided into six steps: first, the screening of valuable stocks. The screening result is relatively in line with China's actual situation; the second step, the selection of principal components indicators by principal component analysis; the third step, the construction of multi-layer feedforward neural network model, including the design of network topology, the modified BP algorithm selection, and selection of the training function; the fourth step, the implementation of multi-layer feedforward neural network by MATLAB. The empirical results show that the non-linear model based on financial indicators can predict the stock price well. The accuracy rate of directional forecast can reach 62.1%. It indicates that the value investment philosophy applies not only in the United States and other countries, in China, we can continue to use this type of theory and share huge benefits under the guidance of it; fifth step, the construction and implementation of the linear estimation model; the sixth step, the comparison of the two models, the result shows that multilayer feedforward neural networks in stock price fitting forecast model is superior to linear estimates model.