Abstract
There are two general approaches to stock price prediction – fundamental analysis, which considers company, industry, and general economic information; and technical analysis, which relies solely on historical price trends. While most portfolio managers champion fundamental analysis, advances in technology have opened up new and more efficient methods for technical analysis.
One of which is the use of LSTM networks to predict future values from time-series data. This study chose 15 PSEI stocks and trained an LSTM model for each using 17 years of daily stock price data. To complement this, we also developed a simple portfolio management algorithm that buys and sells stocks based on the predicted prices. Overall, our implementation outperformed the benchmark Buy & Hold strategy for a portfolio with equal proportions of the same 15 stocks.
