The goal of this project is to train myself to build database and apply machine learning techniques to build a trading strategy on one stock BHP Billiton, an Austrian mining company. The intuition of this work was to assess whether confounders such as the prices of the extracted materials are good predictors of the stock's ups and down, and if a trading strategy could be put in place based on that. I chose yfinance as the free API that has a lot of data and is rather easy to use to get the data.
Limitation:
- computing power to optimize hyperparameters.
- simplicity of variables, some specialized trading parameters could be missing and would improve the code
- lack of sentiment analysis parameter (the SIX and Gold indices are proxies, not directly built through convolutional, RNN or NLP).
- data availability: China's economy is a big part of the demand and its' economic's health is not accessible on yfinance before the 2000, we could use World Bank data or FED data to remedy this.