AN ENSEMBLE METHOD FOR SENTIMENT ANALYSIS ON TEXTILE DATASET USING MACHINE LEARNING ALGORITHMS

Abstract

Sentiment analysis has become a vital technique in today's environment, greatly aiding in the comprehension of user attitudes in product reviews. This research explores sentiment analysis as it relates to reviews of E-Commerce products, primarily concentrating on polarity detection. A whole preprocessing pipeline is covered by the research, which includes actions like resolving missing values, removing symbols and punctuation, converting text to lowercase, deleting stopwords, stemming, and tokenization. After undergoing various preparation methods, the count vectorizer idea is used to convert the dataset into a numeric representation. Then, two popular machine learning techniques—Naive Bayes and Support Vector Machine—are used to determine which reviews are polarity-based and evaluate each algorithm's performance. Furthermore, an ensemble model is suggested that combines the Random Forest and XGBoost algorithms to improve polarity identification performance and accuracy even more. With the purpose of delivering a comparative analysis to support the creation of consumer purchase decisions and product enhancement plans, the study seeks to shed light on the efficacy of these algorithms in the context of E-Commerce sentiment analysis.