Strategies of stock market using prediction of neural network in comparison with models autoregressive

André Pacheco Miranda, Paulo Sergio Ceretta, Luis Felipe Dias Lopes


When investors decide to venture into stock markets they search for a method that provides reliability to support their decision making process even though there is no means to know empirically with certainty which stocks will become profitable investments and which prediction method is the best to discover this. This article presents the development of a heuristic method that employs the trading volume and return lagged as transitions variables. Using assessment criteria proposed further, the result obtained by the development and application of the neural network is compared with the prediction extracted from a linear model. The heuristic method, that has a neural network multilayer perceptron trained with an algorithm for back propagation error, is then compared with an autoregressive moving average (ARMA) model. The results point out that the neural network offers a greater explanation power than ARMA models, even though neither approach presents a satisfactory performance.



Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 International License.






Revista de Administração da UFSM. Brazilian Journal of Management

Universidade Federal de Santa Maria, Santa Maria, Rio Grande do Sul, Brasil, eISSN 1983-4659