Portfolio Allocation of Stocks in Index LQ45 using Deep Reinforcement Learning

Portfolio allocation is an investment strategy in which investors determine the weight for each stock in the portfolio. Using a portfolio, an investor can manage the return and the risk of stock investments. Many methods have been developed to manage a portfolio. One of the most recent is Deep Reinforcement Learning (DRL). In this paper, DRL is applied to construct a portfolio. The portfolio consists of stocks in the LQ45 index in the Indonesian Stock Exchange. The data used is daily closing price data from January 15, 2014, to January 1, 2020. The experiment is conducted, including a combination of the number of shares in the portfolio 3, 5, 7, and 42 stocks. Our results show that the portfolio value and the Sharpe Ratio of the DRL portfolio are better than the Equal Weight and Mean-Variance portfolio. Also, the performance of the DRL portfolio is much better for a small number of stocks.