Artificial neural network: An alternative in forecasting the Philippine Stock Exchange Index (PSEi)

This paper aimed to explore the application of Artificial Neural Networks (ANN) as an alternative tool in forecasting the returns of the Philippine Stock Exchange Index (PSEi) from 2008 to 2015. ANN is a computational model that simulates the structure and function of human biological neural networks. The end-product created by this research was a neural network named PSEi_NeuralNet which was built from a framework called Neuroph. The PSEi_NeuralNet was used to predict the daily PSEi levels through short-term (5-day, 10-day input), medium-term (30-day input), and long-term (90-day, 120-day) input forecasts. Samples were drawn from 2008 and 2015 to forecast the Philippine market during financial crisis and normal times. The Error, Percent Error and Accuracy were used to measure the out-of-sample accuracy of the software. Results showed that forecasts for 2015 had a high accuracy level. However, paired t-test showed that only the 5-day and 90-day inputs have no significant difference with the actual PSEi levels. Long-term forecasts for 2008, using 60-day training data, showed low accuracy results which means that artificial neural networks does not perform well during times of financial crisis. However, short-term forecasting for 2008 gave a highly accurate result