I have a few questions on how to clean/preprocess financial time series data. Are there any sources that can help with this?

The questions are:

  • How should I adjust the data to deal with price changes related to dividends, stock splits, buybacks, etc?
  • How should I deal with outliers? Should I remove them or keep them since they produce important information?
  • How should I detect and deal with missing or otherwise corrupted/incorrect data?
  • How does my cleaning methodology change based on my needs, i.e. should I clean differently for risk management vs. volatility forecasting?
  • How should my procedure change based on the frequency of my data?
  • How can I denoise/improve the signal-to-noise of my time series for better parameter estimation?
  • When should I avoid different data augmentations/processing? What are the drawbacks of each augmentation?
  • What other considerations am I missing?

Basically, is there a comprehensive (or as comprehensive as possible) guide to cleaning financial time series data that includes these topics? I'm mostly interested in data dealing with equities, indices, and derivatives on these.

I also read somewhere about using AI for this, are there any good resources on that?



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.