Investment Model Based On Alpha Value Prediction Research

In recent years, the quantitative investment technology is more and more hot in the domestic financial market. Quants hope to be able to get excess returns by carefully choose stock portfolio. Therefore, this article is proposed an investment model based on stock Alpha value prediction, under the active portfolio management theory, combined with the capital asset pricing model(CAPM) and the financial time series modeling method.First of all, in order to obtain the history of each stock’s excess yield Alpha, this article choose the history of the 300 and the csi 300 index’s day return ratio to build regression model, based on the capital asset pricing model formula. And to do stock Stationarity and White noise test to each Alpha sequence, the results show that most of the Alpha sequence is non-stationary and has strong autocorrelation connections. So it is right to use the time series method to analysis stock’s Alpha sequence.Second, this article build two kinds of models, the univariate and multivariate time series prediction model.And we have test the model prediction accuracy and stability,the results show that the accuracy of monthly average of two kinds of model prediction is 88%.Then, we choose stocks who has a positive Alpha prediction value to build an alternative stock pool.And choose stocks from the alternative stock pool to build stock portfolio.Finally, this article design an entire investment trading strategies according to the building of the investment portfolio.And the investment portfolio has annualised returns of 58% and maximum retracement of about 11% during the empirical.What’more, as compared with the benchmark csi 300 index, the contrast results show that the portfolio strategy completely have the ability to beat the benchmark. It illustrates the effectiveness and practical significance of the investment model in this article.