How do the naive Bayes classifier and the Support Vector Machine compare in their ability to forecast the Stock Exchange of Thailand?

TL;DR

This study compares naive Bayes and support vector machine classifiers in forecasting the Thai stock exchange, finding naive Bayes slightly more effective but both are limited in accuracy.

Contribution

It provides an empirical comparison of naive Bayes and SVM for stock market prediction using real data and standard evaluation metrics.

Findings

Naive Bayes outperforms SVM in prediction accuracy.

Both algorithms have limited effectiveness in modeling the stock exchange.

The study uses real market data and standard error metrics for evaluation.

Abstract

This essay investigates the question of how the naive Bayes classifier and the support vector machine compare in their ability to forecast the Stock Exchange of Thailand. The theory behind the SVM and the naive Bayes classifier is explored. The algorithms are trained using data from the month of January 2010, extracted from the MarketWatch.com website. Input features are selected based on previous studies of the SET100 Index. The Weka 3 software is used to create models from the labeled training data. Mean squared error and proportion of correctly classified instances, and a number of other error measurements are the used to compare the two algorithms. This essay shows that these two algorithms are currently not advanced enough to accurately model the stock exchange. Nevertheless, the naive Bayes is better than the support vector machine at predicting the Stock Exchange of Thailand.