ML Investing (Final Project)

This project serves as a guide to the some of the focal points in a ML investing process. The two main parts that the project goes into detail are Portfolio Management and Stock Prediction.

Stock prediction is done tradtionally through regression. However, this project delves into more recent methods that've shown progress such as LSTM RNN models, and Sentiment Analysis. From the results, I believe we could improve the project by gathering more data based on satalite imagry to predict quarterly returns for physical stores. I would also try obtaining more technical analysis features to improve neural network and Sentiment analysis peformance.

Portfolio Management is usually done through a means of linear optimization for optimal weights. I decided to try not to look at pure optimization but investigate the huge variance in performance of stocks in different macroeconomic shifts and focus on ones with lower variance. This led to PCA and it turned out to be a very optimal tool for finding a good portfolio in times of macroeconomic shifts. I also intend to try Monte carlo simulation as well to see if I get other results and compare them.