Multi-day Stock Price Prediction Based On GA-BP Neural Network

In this paper,a GA-BP neural network model is established for predicting multi-day stock prices.BP neural network can be used to find out the internal law of stock market volatility based on stock historical data to predict stock price changes.But improper initial weights may cause BP neural network to fall into local minimum and lead to low prediction accuracy.Because genetic algorithm is a heuristic global search algorithm that is not easy to fall into the local optimal trap in the search process,it can be used to optimize the initial weights and thresholds for BP neural network to improve its prediction accuracy.The current research on stock price forecasting basically uses multi-day data to predict the price of the next day,while the Chinese stock market adopts a trading system of T + 1,that is,investors can not sell stocks on the day they buy stocks,at least until the next trading day after buying stocks.Most investors wonder whether they should buy stocks or not because they have no idea about the stock price trend on the next day.In this paper,the GA-BP neural network model is constructed to predict stock prices for several consecutive days.Firstly,the genetic algorithm is used to obtain the optimal initial weights and threshold values with three consecutive days of opening price,maximum price,minimum price,closing price,and turnover as input variables,and then BP neural network is constructed to predict the next two consecutive days of opening price,maximum price,minimum price,and closing price,the prediction error is less than the error of the model others studied.Empirical analysis shows that the investment risk of the model is larger than that of Shenzhen Changcheng Investment,but the accumulated return of the former is more than 50% higher than that of the latter.The second is to extract the principal component of 11 commonly used indicators and then use the principal components of four consecutive days as the input variables to obtain the optimal initial weights and threshold values using genetic algorithm.Then the construction of BP neural network predicts the opening price,the highest price,the lowest price,and the closing price for the next two consecutive days.The empirical analysis shows that the sharp ratio of simulated investment is larger than that of Shenzhen Changcheng investment.It shows that the investment strategy given by the model is less than the investment risk of Shenzhen investment,and the accumulated return of the former is more than 50% higher than that of the latter.In a word,the prediction model established in this paper has higher forecasting precision and reference value.