Sentiment Classification of Nykaa Product Reviews Using Classical Machine Learning Models

This study presents a machine learning pipeline for the sentiment classification of Nykaa cosmetics product reviews. The research utilizes a dataset of 61,271 reviews, categorized into three sentiment classes (Negative, Neutral, and Positive) based on user star ratings. The methodology encompasses text preprocessing (lowercasing, regex-based cleaning, and stop-word removal) followed by feature extraction using TF-IDF, restricted to the 5,000 most frequent terms. Six classical supervised classifiers from scikit-learn are implemented and evaluated. The paper provides an exploratory analysis of review distributions across brands and sentiment classes, addressing common challenges in real-world e-commerce data such as class imbalance and multilingual "Hinglish" text. The study concludes by outlining future directions, including aspect-based sentiment analysis for specific product attributes (e.g., fragrance, packaging, longevity), hyperparameter tuning with cross-validation, and the exploration of transformer-based models like BERT.