When creating the input features for time-series data for a particular stock, I understand that it would be beneficial to use fractional differencing to ensure that the Close Price is stationary. It can then be used as one of the input features.
If i would like to include technical indicators, such as RSI, ADX, etc (using the TA-Lib package) as additional input features, should I use the Fractionally Differenced Close Price or the Raw Close Price in order to derive the value of RSI, etc?
Any clarification is appreciated.
Thank you.
Technical indicators tend to move in a range. For example, the RSI indicator can range from 0 to 100. Therefore, it is not necessary to create technical indicators using the differenced price. However, you can try using raw price and differenced price to calculate the technical indicators. And pass both features separately to ML algo and analyse if there is any improvement in the prediction.