how to Choosing Machine Learning model for trading strategy
This is an interesting question which has come up in every ML practitioner's mind at least once in their life. Before we answer this, you should note that machine learning is in a way trying to mimic the human way of thinking and add a computer's blazing fast processing speed to it. That being said, numerous ML models are available, but each was created to solve a different problem. If you have a dataset which already has a label (think pictures of cats and dogs), you would use supervised learning. If you are designing ML to cluster similar assets together (based on fundamentals), you could use unsupervised models.
Ultimately, it comes down to your choice. With respect to the trading domain, you will find that starting with a simple ML model (decision trees, support vector machine) can give you a decent performance on backtests. With the help of python language, it has become considerably easier to create and test different ML models with the help of ML performance metrics such as accuracy, and confusion matrix(for label specifics). You can always create a few machine learning models, analyse its performance. And then move forward with the one which is delivering better results. Of course, my recommendation would be to first paper trade your strategy and once you find the performance satisfying, move forward and live trade.
Hope this helps.
Thanks … current i am learning decision trees, support vector machine and read many quantra blogs .