Predicting Customer Sentiment from Bangladeshi E-Commerce using Machine Learning Techniques

E-commerce or electronic commerce is a term used to describe the various business activities and transactions conducted on the internet. It has revolutionized the way business and consumers interact with each other in buying goods and services commonly used today. E-commerce involves different online deals, including retail shopping, banking, investments in stock markets and ample of other basic commercial procedures that the consumers can move from their comfort zone. This study checked the level of shopper fulfillment on e-commerce in Bangladesh by product reviews. This was done by conducting it in a dataset of 6964 events classified into three categories "satisfied", "not-satisfied", "highly satisfied". Based on the wide variety of models as presented, this research paper iterates using SVM, LR and random forest among other standard Machine Learning Techniques i.e., Standard Machine Learning technique (SVM), Logistic Regression (LR), Random Forest, Decision Tree, Gradient Boosting Classifier etc. to discover and predict customer sentiments. The Random Forest (RF) classification achieved a better result than standard, with 94% accuracy in learning complex patterns of Bangla product reviews and predicting customer satisfaction effectively. Transparency and ethical aspects of methodology guarantee validity and generalization. Further research, therefore on long-term analyses, international comparison and new technologies are needed to explain the structure of customer satisfaction in the changing e-commerce environment especially in Bangladesh. This study will be helpful for the investigators, developers and politicians who want to enhance user experiences and methods in the competitive growing e-commerce market of Bangladesh.