We can make further improvements and add additional features.
- Utilize ensemble learning on LSTM model
- Use different historical stock price dataset sources other than Yahoo Finance API
- Use other time-series model to predict the stock price other than ARIMA (e.g. Stochastic Volatility Model)
- Add additive decomposition function to compare with the multiplicative decomposition
- Construct other portfolio for evaluation (e.g. Markowitz’s mean-variance portfolio)
- Utilize Robust Optimization to get the covariance matrix of the stock log returns
- Add other risk distribution strategy to get the budget
- Use other risk measures for the risk parity portfolio (e.g. Conditional Value-at-Risk (CVaR))