Bengali Sentiment Analysis of E-commerce Product Reviews using K-Nearest Neighbors

Abstract

The sentiment analysis of the Bengali language is converting into a trendy research topic nowadays. Sentiment analysis is a useful technique in opinion mining, emotion extraction, and trend predictions. By sentiment analysis, the actual sentiment of a text review can be extracted. Every day, every second's people use the internet for different purposes and leave their opinions or perspective views in various places on the internet as a text format. The opinion or review on the internet can contain the author's positive, negative, and neutral views of the statement. This study proposed a machine learning-based model to predict a user's sentiment (positive, neutral, and negative) of a Bangla text review. We have applied five machine learning algorithms in our dataset, which we manually gathered from a Bangladeshi e-commerce site called “Daraz.” We have experimented with Random Forest classifier, Logistic Regression, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and XGBoost algorithms with our dataset. KNN performs great among all these five algorithms in all the performance measures of accuracy, precision, recall, and f1-score. KNN shows 96.25% accuracy, 0.96 in each of precision, recall, and f1-score.