Enhancing the Performance of Sentiment Analysis Models Using GridSearchCV: A Case Study on Electric Vehicles in Thailand

This study investigates the enhancement of sentiment analysis model performance through hyperparameter tuning using GridSearchCV, with a focus on electric vehicle reviews in Thailand. To address the challenges inherent in Thai language processing, PyThaiNLP was employed for word segmentation and text preprocessing. Four machine learning models-Support Vector Machines (SVM), Multinomial Naive Bayes, Logistic Regression, and Stochastic Gradient Descent-were implemented for sentiment classification. The hyperparameters of each model were systematically optimized to determine the configuration that maximizes accuracy, precision, recall, and F1-score. Experimental results revealed notable improvements across all models following optimization. The SVM model, identified as the best-performing classifier, achieved an accuracy increase from 75.00% to 76.69% and an F1-score improvement from 74.82% to 76.69%. The optimal SVM configuration employed a radial basis function kernel with a regularization parameter (C) of 10 and a gamma value of 0.1. These findings underscore the significance of hyperparameter optimization in improving model effectiveness and contribute to advancing sentiment analysis in linguistically complex environments such as Thai.