Hybrid Machine Learning Methods with Malay Lexicon for Public Polarity Opinion on Water Related Issue

Opinion classifications from Twitter are still in demand among research works on related opinions or feelings expressed on various issues. One of the concerns expressed in Twitter is on water-related issues such as the lack of clean water supply. It has been found that the issue highlighted in Twitter is the frequent disruption of clean water supply in Malaysia. The discussions concerning this issue contain positive and negative emotions like anger, joy, worry, and frustration. The focal point of this article is to evaluate hybrid sentiment analysis using a machine learning classifier to analyze the polarity of opinions employing real data from Twitter. A series of experiments were performed on a hybrid of deep learning, support vector machine, Naïve Bayes and random forest with a lexicon-based model. In addition, the Malay sentiment lexicon score is proposed. The Malay sentiment lexicon scores have improved the accuracy and F1-score of all hybrid methods. The analysis uncovers that negative and positive polarity opinions can be beneficial to the relevant authorities to overcome the water supply disruption issue.