A comparison between SVM and multilayer perceptron in predicting an emerging financial market: Colombian stock market

Abstract

Achieving accurate stock market forecast impacts strongly to investors, like retirement funds and private investors, giving them tools for making better data based decisions. This article studies the applicability of two soft computing methods, Artificial Neural Networks and Support Vector Machines, to forecast Colombian stock market. Technical indicators were selected as inputs of the machine learning techniques, and up/down movement was selected as output. Cross-validation was employed to improve generalization, and automatic parameter tuning was performed to improve model performance. The results showed that Support Vector Machines performance was better than Artificial Neural Networks, and the results are similar to those found in other studies.