Backtesting is done by looking at historical user tweets over a range of trading days. Theoretical daily returns are then calculated based on the most confident daily selected stocks. Below is an overview of the general system and data flow before backtesting begins.
The main steps that must be done before backtesting are locally saving stock price data and generating stock features based on pre-computed user features.
- 1.Save historical stock open-close prices
- 2.Generate user features to be used in weighting user tweets
- 3.Generate daily stock features based on user features
- 4.Backtest using generated stock features