We propose an innovative but simple method to predict the monthly returns of individual stocks utilizing solely past and publicly available information. This technique adopts the concept of the person-centered approach that has been widely used in psychological analysis of individual human behavior. Our approach employs underlying hidden correlation structures among all relevant factors or predictors available, which is different from the conventional studies that mainly focus on identification of certain factors or predictors themselves. Choosing a generic set of financial ratios as predictors, the portfolios constructed based on our method produce economically sizeable monthly abnormal returns up to 1.81% after incorporating transaction costs in various asset pricing models.