As the number of e-commerce sites has increased over the years, there has been a lot of data generated through product reviews. The reviews act as an important source of information on how customers perceive products, their likes and dislikes, and overall satisfaction. Analyzing this massive a mount of textual data manually takes time and is not efficient and prone to errors. In order to solve this problem, the proposed project focuses on building a system that can automate sentiment analysis by categorizing customer reviews based on their polarity, which includes positive, negative, and neutral, along with the use of emojis.The proposed system employs different NLP techniques to preprocess the raw textual data. Some of the techniques include text cleaning, tokenization, stop-word removal, and emoji transformation. Once the raw data has been cleaned, the TF-IDF technique is used to extract features from the textual data and convert them into vector form. In the case of sentiment classification, Logistic Regression model is utilized owing to its effectiveness and applicability in text classification. The classification is conducted via training and testing of the machine learning model on a set of data consisting of users’ reviews. In addition to traditional evaluation methods like Accuracy, Precision, Recall, and F1-Score, a hybrid method involving rule-based correction is adopted.From experimental findings, the suggested method has shown satisfying performance and obtained around 80-85% Accuracy level. It can be seen that adding emojis to the input improves sentiment recognition since emojis have high levels of emotionality. Positive and Negative sentiments are easily recognized by the system, whereas Neutral Sentiment Classification seems to be more difficult. In conclusion, this project demonstrates that the use of machine learning along with natural language processing approaches can effectively be used for sentiment analysis. The suggested approach has been found to be efficient and scalable enough to analyze massive amounts of data generated by users.

This research developed a sentiment analysis model that uses machine learning algorithms and natural language processing to classify text reviews as positive, negative, or neutral. The addition of emoji analysis greatly improved the accuracy of sentiment detection, showing its importance in text analytics applications. When comparing the two algorithms, Logistic Regression showed better accuracy, while Naive Bayes was more efficient in its calculations. E-commerce platforms can use this system to evaluate customer feedback effectively.

[1] IEEE – IEEE, “Advancements in Sentiment Analysis Using Hybrid Machine Learning Models,” IEEE Access, 2025. . [2] Springer – Springer, “Natural Language Processing Techniques for E-commerce Review Analysis,” 2025. [3] ACM – ACM, “Improving Text Classification Using TF-IDF and Logistic Regression,” ACM Transactions, 2025. [4] Bing Liu – Liu, B., “Sentiment Analysis: Trends and Applications in E-commerce,” 2024. [5] Springer- Springer, “Text Classification Using Logistic Regression and TF-IDF,” 2024. [6] Elsevier – Elsevier, “Role of Emoji and Text Preprocessing in Sentiment Analysis,” 2024. [7] Bo Pang – Pang, B., Lee, L., “Opinion Mining and Sentiment Analysis: A Review,” 2023. [8] ScienceDirect – ScienceDirect, “Applications of NLP in Product Review Analysis,” 2023. [9] Bing Liu – Liu, B., “Sentiment Analysis and Opinion Mining,” Morgan Kaufmann, 2021. [10] Elsevier – Elsevier, “A Survey on Machine Learning for Sentiment Analysis,” 2020.