Enhanced Malay Sentiment Analysis with an Ensemble Classification Machine Learning Approach

摘要

Sentiment analysis is one of the challenging and important tasks that involves natural language processing, web mining and machine learning. This study aims to propose an enhanced ensemble of machine learning classification methods for Malay sentiment analysis. Three classification approaches (Naive Bayes, Support vector machine and K-Nearest Neighbour) and five ensemble classification algorithms (Bagging, Stacking, Voting, AdaBoost and MetaCost) were experimented to achieve the best possible ensemble model for Malay sentiment classification. A wide range of ensemble experiments are conducted on a Malay Opinion Corpus (MOC). This study demonstrates that ensemble approaches improve the performance of Malay sentiment-based classification, however, the results depend on the classifier used and the ensemble algorithm as well as the number of classifiers in the ensemble approach. The experiments also show that the ensemble classification approaches achieve the best result with an F-measure of 85.81%.