Indian Journal of Science and Technology

Abstract

Objective: To examine the sounds in Thai videos on YouTube, to analyze consumer opinion on beauty products. Statistical Analysis: The data was collected to analyze the sentiments of Thai sounds from 500 YouTube videos which were the reviews of beauty products; the length of each video being approximately 2-5 minutes. Findings: The accuracy rate of SVM (SVM) appears greater than those from the Naïve Bayes (NB) and K-Nearest Neighbor (KNN) techniques. The SVM used the RBF Kernel-typed Sequential Minimal Optimization (SMO) function, where c=50000 and gamma=0.1; the accuracy rate was 94.40%, when using K-fold Cross Validation, where K =10, that had 293 attributes. Application/Improvement: The SVM used the RBF Kernel-typed SMO function, where c=50000 and gamma=0.1; it can be applied to analysis of sentiments studies in social media derived from Thai videos which are evaluate processes need to be fast and able to provide negative, neutral, or positive results in a timely manner, in which it will become a purchase decision-making guide for consumers.

Keywords: Sentiments Analysis, Social Media, Support Vector Machine (SVM), Thai Sound, Video