Sentiment Analysis for Informal Malay Text in Social Commerce

Abstract

Sentiment Analysis (SA) is opinion mining which often defines as the study of emotions, opinions, or feedback that relates to the usage of computational linguistics, text analytics, and natural language processing. With the rise of social media posts, it is becoming more challenging to evaluate brief, casual, and non-structured texts to optimize consumer feedback and spot patterns. Meanwhile, social commerce involves social media for social interaction in assisting customers and merchants to do business transactions. From a social media perspective, the informal Malay Text is less explored by the researchers. Thus, it will directly yield difficulties in conducting and preparing the SA processes. Cross-Industry Standard Process for Data Mining (CRISP-DM) was adapted as a reference model for the methodology of this work with machine learning approaches in classifying the informal Malay textual data based on sentiment. The dataset was extracted from the Facebook platform of Pos Laju Malaysia pages. The comparison of the classification technique performances was analyzed in identifying the most accurate classifier for SA, within three different machine learning classifiers was experimented by using 1200 instances from an informal Malay textual dataset. The results of Decision Tree (J48), Support Vector Machine (SVM), and Naïve Bayes (NB) were analyzed and discussed. The result of the highest accuracy of Ten-Fold Cross-Validation is 69.7% and meanwhile, for the Percentage Split method, the highest accuracy result is 70.9%. It shows that Support Vector Machine (SVM) is the best classifier compared to other classifiers of text classification based on sentiment.