Neural network ensemble for financial trend prediction

Presents a study of an artificial neural network that works collectively as an ensemble for the task of predicting the future trends of a financial time series from a classification approach. The Kuala Lumpur Stock Exchange composite index is chosen as a case study. A multilayer perceptron neural network is trained by backpropagation algorithms to capture the relationship between the future trends and the past values of the index. The performance of the neural network ensemble is compared to the single neural network. A confidence measure using agreement (variance) within the ensemble and with the ensemble output is studied.