Using the MLP Classifier Model with Markov Chains Observing the Bovespa Index

Resumo

The application of resources in the financial market in the form of investment is primarily linked to variable income, which has a higher risk, requiring a more thorough assessment to increase its assertiveness. Successful past experiences influence decisions and can lead to correct choices. Mathematical and statistical models should assist in decision-making to optimize their success. This work involves creating a web service connector to capture data from the BOVESPA index, randomly used to choose between three websites: Infomoney, Investnews, and Valor Invest. This work used an Artificial Neural Network (ANN) Classification model, the MLPClassifier, and compared this model alone and about itself with the insertion of the modeling with Markov Chains. The MLPClassifier model proved satisfactory, given that performance metrics measure its classifications, and the speed will depend on the number of hidden layers used. By including Markov Chains in the MLPClassifier model, it was possible to refine the classification process of the model, which already had a very high level of assertiveness.