Situational Portfolio Forecasting and Allocation with Deep-Learning Approach

Abstract

Portfolio optimization is selecting the best set of possible weights for a group of assets where the objective is to maximize the returns and risk-return ratio and minimize the risks and volatility. This research aims to develop and test ARIMA and LSTM as forecasting techniques and subsequently perform portfolio optimization using a custom optimization methodology leveraging the forecasted returns from the models mentioned earlier. The intention is to develop a portfolio that dynamically allocates weights to the assets for the optimum investment strategy. The portfolio considers an initial investment of 100 units of currency, allowing uncomplicated interpretation of results and data.