Developing research-based earnings forecasting models using data scraped from Compustat, IBES & WRDS.
The end goal is to identify 3 high growth potential stocks, on the basis of the least biased and most accurate earnings prediction model.
The following features were extracted/engineered:
- earnings indicators
- accruals (control for earnings quality)
- total assets (control for size)
- income/balance sheet variables
- lagged variables
- Text-based features extracted from MD&A section
I also leveraged research-based models such as:
- HVZ
- Residual Income
- Earnings Persistence