I am confused with the useage of the concept "Alpha Model" in quantative investment. According to Qian, Hua & Sorensen (2007), the first thing in the toolbox of quantative investment process is "an alpha model that forecasts excess return of stocks"(Page 5). And on page 81, the author mentioned that "an important component of any successful investment strategy is forecasting expected returns using alpha models".
So, what should an alpha model do? Predict excess return or predict expected return? Given the word "alpha" in the term "alpha model", shouldn't it predict the "alpha"(i.e., the risk-adjusted return, e.g., Jensen's Alpha) for each stock so that we can select stocks with high alphas to construct portfolio?
In addition, a common method to evaluate an "alpha factor" is to calculate the "information coefficient", which is usually defined as "the correlation between the forecasts and the eventual returns"(Grinold & Kahn, 2000; Qian, Hua & Sorensen, 2007). But an "alpha factor" itself is just a "number" calculated for each stock(see this post), not a return forecast, and some people seem to just use the correlation between the raw alpha factor values and stock returns as the "information coefficient". So I wonder which is right for the calculation of information coefficient. If we calculate it as "the correlation between the forecasts and the eventual returns", then how to get the "forecasts" given the raw alpha factor values?
Grinold & Kahn, 2000, Active Portfolio Management: A Quantitative Approach for Producing Superior Returns and Controlling Risk
Qian, Hua & Sorensen, 2007, Quantitative Equity Portfolio Management: Modern Techniques and Applications