Bangla E-Commerce Sentiment Analysis Using Machine Learning Approach

Abstract

In this paper, a machine learning approach is used to predict user sentiments from Bangla texts about products available on e-commerce sites. In order to accomplish the task, we have constructed a Bengali corpus of the public views about products and services of multiple Bangladeshi E-commerce organizations. Besides, we have applied six different machine learning algorithms (Multinomial Naive Bayes (MNB), Logistic Regression (LR), Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), and Stochastic Gradient Descent(SGD)) to predict and analyze the polarity of public sentiments. Term Frequency–Inverse Document Frequency (TF-IDF) technique has been applied by using Trigram features. Finally, after optimizing the hyperparameters using the Randomized-SearchCV algorithm, SVM classifier has been found to demonstrate the highest accuracy of 90.68% for predicting public sentiments.