Hello dear community.
I am completely new to the topic of stock market prediction.
I understand the concept of Numerai and have worked as a Data Scientist for a tech company for multiple years. My specialties are RNNs, so I know a bit about time series.
I already build a classical LSTM model and let it run on the medium numerai dataset, however the results are not great.
Now I am asking for any kind of tips to get better.
As far as I understood the most popular model architectures are XGboost, Transformer and RNNs, right?
Is there any github repo for a model that is performing decently?
Any other ressources that you can recommend?
Thanks in advance!
I can’t tell you what you’ll consider useful, but here’s what I considered useful, when I started (which was 4 months ago ):
- I think you should go trough the Numerai Example Scripts, that would guide you trough the whole process of training, submitting, etc.
- Then Check the Benchmark models of Numerai, this is really fresh information: Benchmark Models - Numerai Tournament
- Watch the Numerai content on YouTube, specifically the Quant Club and the Fireside chats to give you better context.
- Watch StudyM8’ts Numerai Series
- Check the Grid Search article: Super Massive LGBM Grid Search and in general, just search the forum if you happen to have questions
- Just hop into Discord, a lot of smart people discussing so many things, I picked up a lot there
- Marcos López de Prado: Advances in Financial Machine Learning book will give you so many new concept and basic understanding on what Numerai does or even Finance
As per what models are considered effective?
I can’t really tell, most people are using GBT, there are a few people active on Discord who are using, Genetic Algorithms, Random Forests, Transformers, RNNs, but not much people are conversing about LSTMs.
Amazing, thank you. That helps a lot