Chaos Theory Enhanced LSTM Model of the Philippine Stock Exchange Index

This study advances the field of financial time series forecasting by merging chaos theory with machine learning, particularly through the use of long short-term memory (LSTM) networks, to predict the Philippine Stock Exchange Index (PSEi). Recognizing the limitations of traditional linear models, this research adopts Takens’ embedding theorem to reconstruct the phase space of PSEi, enabling a profound comprehension of its intricate, nonlinear behaviors. Through rigorous data preprocessing—including multiple imputation by chained equations for missing value treatment—the study facilitates a reliable analysis platform for LSTM application. The performance of the Takens’-LSTM model, gauged against actual market data, demonstrates its ability to capture the overall stock price trend with high fidelity, despite observed deviations during volatile market periods. This exploration not only contributes to theoretical knowledge on financial forecasting but also introduces a practical forecasting tool, potentially benefiting a wide spectrum of market participants.