Artificial Intelligence And The Multivariate Approach In Predictive Analysis Of The Small Cap Index Of The Brazilian Stock Exchange

Abstract

Looking for greater returns, the Brazilian investors have increasingly resorted to the stock exchange. In this way, companies classified as Small Cap appear as a great opportunity for significative profitability given the higher volatility. In this scenario, studies aiming to explain the behavior of the stock market are becoming even more important. This article aims to develop prediction models for the Small Cap Index in order to assess whether the series under study can be explained by economic and financial variables. The monthly data used were collected from October 2005 to December 2019. The methods applied for the development of predictive models were multiple linear regression and the multilayer perceptron artificial neural network with backpropagation supervised learning algorithm. For the first forecasting method, the model presents, based on 10 predicting variables, approximately 91% of the explanatory capacity of the Small Cap Index variability, against approximately 98% of the second method, which is based on 16 variables for the forecast. Although the artificial neural network presents better prediction results, both models are satisfactory to explain the behavior of the index under study.