Abstract of Thesis presented at COPPE/UFRJ as a partial fulfillment of the requirements for the degree of Master of Science (M.Sc.)

Artificial Neural Networks in the Main Ibovespa Index Series Prediction and its Application on Automated Trading Systems

Igor Ramalho Pommeranzenbaum

March/2014

Advisor:

Luiz Pereira Calôba

Department:

Eletrical Engineering

      The application of techniques in financial time series prediction is a subject of constant and high interest in the community, both for investors and researchers. It is a challenging area in relation to the complexity of problems and can generate high financial returns for companies involved with asset analysis and automated trading on stock exchanges. In this work, we present a model for predicting future values on the main series of Ibovespa index - namely close, high, low and order, which means the order that the high and low occur - using Artificial Neural Newtorks (ANNs). For inclusion in the model, various time series of global market indexes were obtained, being submitted to classical time series pre-processing methods, so that could be selected as inputs to the ANNs by a cross-correlation criterion. Finally, a simulation environment was created to perform the application of the results obtained by the ANNs using different market strategies, better known as Trading System. The output of this simulator is a set of numerical indicators that show how would be the results for a hypothetical financial market application in the real world.