1. Indicator selection and stock return predictability, by Zhifeng Dai , Huan Zhu
In this research paper, the authors tried to solve the problem that how to build a winning portfolio from previous simple indicators/strategies. The method that the authors used is momentum-determined indicator-switching (MDIS) strategy.
Though this research is applied to stock market return, and the predictors include micro- and macro-variables, but the idea is worth to try to apply to stock prediction.
2. Forecasting stock prices, by Arie Harel, Giora Harpaz
In this paper, the authors group the stock to three types: overpriced stocks (OP), underpriced stocks (UP), and fairly-priced stocks (FP). Then, the authors show how to calculate evaluation metrics for a prediction model. The metrics are quite popular to machine learning practitioners (sensitivity, specificity, ROC curve). Whatever, the idea about classifying stocks to OP, UP, and FP is interesting and is worth to deeper dive-in.
Wish that you can find some ways to modelize your trading objectives based on the idea of this paper.
3. Four centuries of return predictability, by Benjamin Golez, Peter Koudijs
In this paper, the authors test the predicability of stocks’ annual returns by dividend yields.
The paper shows that the dividend-to-price ratio tends to increase before a period of high returns and vice versa. Otherwise, the return predictability from dividend yields is related to business circle, i.e. both the dividend-to-price and subsequent returns tend to increase in recessions.