I want to use a model to trade in finanical market.

which i have several features, like macd, rsi, or other common features.

and my target is to make a tradeable predict in every time point.

so my target can be:

  1. yield in a fixed time laster, like, 30 min. yt = close(t+ws) - close(t)
  2. futures price direction, which only can be 1(price up in the future) -1 (price down in the future)

these are difference between regression and classification.

which one you think is better, and, any suggestions about this problem?



From my personal experience, I think what matters the most in terms of return is how good your risk management is. You can use both regression or classification, but all approaches have some errors associated with it, so in the long run, it comes down to how you can manage incorrect predictions. So you can start with any model, say you go for classification and you achieve 70% accuracy. then you use your model to predict on some historical data (that for sure you have not used during trading) and you build some trading strategy using your forecasts. Then, you can use something like Pythons Backtrader and check how are you doing in terms of return on historical data. Sometimes its not enough to correctly predict price movement in 9 out of 10 cases, because your one loss can be bigger than all the profitable bets and you also need to take into account fees that will burn your profit. Thats a lot of issues you need to think about before putting this model into production. And even using this approach you should be aware that historical success of your model+strategy does not guarantee that it will be the same in the future.

To sum it up: I think it comes down to the strategy and not the model itself.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.