Classification and visualisation of Malaysian courier services based on X sentiment analysis using Naïve Bayes

Users express their views on various topics through social media, including experiences with courier services. Getting insights from such a huge volume of unstructured textual data is scientific challenging. This paper discusses the perception of the Malaysian courier companies J&T and DHL. Sentiment analysis helps businesses determine customer’s opinion and how to improve service quality, but personalized studies on Malaysian logistics are few. This research tackles the challenge of identifying sentiment in X caused by informal language and acronyms. A machine learning techniques Naive Bayes classification model was constructed to tackle this problem with the development process following the Waterfall model. The dataset that was collected underwent tokenization, lemmatization, stop-word removal, and then TF-IDF feature extraction before it was classified into positive neutral and negative sentiments. Performance was improved through hyperparameter tuning on stratified k-fold cross validation training and validation sets. After tuning, the model achieved 86% training accuracy and 78% testing accuracy, hence improving upon the classification performance. The confusion matrix along with precision, recall, and F1-score evaluated the performance of the model. For accuracy in real-time sentiment tracking, support vector machines SVM or KNN could be leveraged. A bigger dataset with aspect-based sentiment might reveal more about service problems. This study contributes to sentiment analysis in Malaysia and demonstrates its applicability in enhancing the logistics customer experience.