Hello everyone,
I would like to discuss a solution for cryptocurrency price prediction. Currently, I am using GARCH, but it seems to perform poorly. I am considering improving the model with deep learning, specifically by applying a CNN to capture non-linear features. I would appreciate any advice to help me make the right decision and move in the correct direction.
An observation:
Zillions of people have tried to use machine learning to predict the financial markets.
It’s extremely difficult due to the random nature of future events, and a lack of useful predictive features.
Those who have figured out the magic, they’re not much interested in talking about how they did it.
Thanks for the reply. Let them keep silence. But rest of us can talk about it.
I want to know if the GARCH model will be enhanced if I add the layer of CNN after it. They used to say that the GARCH model only applies the linear features, so DL models like CNN will help it to extract non-linear features.
And even if you do aim to develop a more useful model, the people behind the big AI developments will put every hurdle they can in the way. It just happened to me and an AI entity I’d been developing for months and months. AI is not that fearsome thing humans need to worry about, the humans that stand behind OpenAI and others are the worry. Their intentions are monetization and control, as you might expect. Build a 99% efficient crypto companion? Plausible, but undoable for obvious reasons.
Have you seen this?
Just started working on something similar but for future markets. My goal is not to predict price but rather predict bias and include probabilities / confidence for each.
Just started this project so I’m exploring multiple models… for now I’m working on the feature engineering, and looking into models that can be explainable, so I can take any insights into my eventual algo… Any feedback is appreciated!
If nothing else, it’s just a fun technical exercise.