The Application Of Time Series Model In Stock Price Forecasting

Stock traders, mutual funds, investors and analysts are hoping to have a way to predict the stock price volatility. Predicting fluctuations of rising (or falling) can help stock traders to make better investment decisions. Good prediction model allows investors to better understand the stock price changes of drivers, to provide support for better risk management.The case is required to predict the Shanghai index day’s closing price, in this paper, we adopt two kinds of forecasting methods. One is using autoregressive and moving average method, to make the appropriate logarithmic differential, find a relatively optimal ARIMA model, and to establish conditions mean prediction model. Another is due to some stock traders (such as stock index futures traders) in predicting the stock price change case, only care about the future stock up or down and don’t care about the specific and values, to facilitate them to do more or shorting. So we combined with the corresponding external data, application for close binary Logistic regression model. And choose a evaluation model of performance standards, make the stock price of the accuracy of the prediction accuracy is better than random guesses.We establish ARIMA model, make ten step forward prediction, and update or correct of the forecasting model, and compare to the real value, compared with the 0.95% average percentage error (MAPE2), deviation is smaller.In a Logistic regression model, we have validation set (1 month 23 days) the data on the prediction, the overall accuracy of the model and the classification matrix statistics for 70%, the accuracy of the prediction accuracy is better than random guesses by 50%, and further puts forward the AUC evaluation method.