Machine Learning — Feature Selection Profitable Alpha Search - Paper Reading

Recently, I’ve been learning about feature selection in machine learning, including filter methods, wrapper methods, and embedded methods such as regularization. It’s been fascinating to study these techniques and apply them to real-world problems, such as predicting the price of bitcoin.

One paper I read recently claimed to have achieved impressive prediction accuracy in the 80–90% range using over 130 features. At first, I was impressed by the results, but upon testing the model on real data, it became clear that the model was overfitted and the accuracy plummeted to just above 50%.

Although the paper was ultimately flawed, it was a valuable learning experience for me. I was able to understand the paper and identify its shortcomings, which will help me in my own work. I will continue to study and master these feature selection methods so that I can replicate and improve upon existing research in the field.

Overall, I feel like I’m getting closer and closer to being able to produce high-quality research on my own. It’s an exciting time for me as a machine-learning enthusiast, and I can’t wait to see what the future holds.