Sentiment Analysis of Customer Product Review Using Machine Learning

Abstract

Understanding customer opinions through sentiment analysis has become essential for businesses aiming to enhance customer satisfaction and refine their strategies. As digital platforms and e-commerce grow, consumers increasingly depend on online reviews for decision-making. This research delves into the analysis of customer product reviews using a range of sentiment analysis techniques, including lexicon-based approaches, traditional machine learning algorithms, and advanced deep learning models like Recurrent Neural Networks (RNN) and Transformer-based architectures such as BERT. The study emphasizes effective text preprocessing methods—such as tokenization, stemming, stopword elimination, and vectorization—to improve classification outcomes. Sentiment classification is assessed using metrics like accuracy, precision, recall, and F1-score to identify the most suitable model. Moreover, aspect-based sentiment analysis (ABSA) is employed to extract detailed opinions related to specific product features such as price, usability, durability, and customer service. The study also addresses challenges such as sarcasm detection, language diversity, and domain-specific interpretations. The insights from this research aim to support the development of intelligent sentiment analysis tools, enabling businesses to monitor feedback efficiently, tailor marketing strategies, and enhance customer experiences for long-term success.

Introduction

1. Introduction

In the digital age, customer reviews across e-commerce platforms and social media provide vital insights into consumer sentiment, expectations, and satisfaction. Sentiment analysis, or opinion mining, is a key technique for extracting emotional tone from unstructured text data. It helps classify feedback as positive, negative, or neutral, enabling businesses to improve products, customer service, and brand loyalty.

2. Sentiment Analysis Techniques

Sentiment analysis involves several stages:

• Data Collection from platforms like Amazon, Yelp, and social media.

• Preprocessing (e.g., text cleaning, tokenization, stopword removal, stemming, lemmatization).

• Feature Extraction using techniques like:

  • Bag of Words (BoW)

  • TF-IDF

  • Word Embeddings (Word2Vec, GloVe)

  • N-grams

3. Machine Learning Models for Sentiment Analysis

Models used include:

• Traditional ML: Logistic Regression, Naïve Bayes, Random Forest, Support Vector Machine (SVM)

• Deep Learning: Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), Convolutional Neural Networks (CNN)

• Advanced Techniques: GPT-4, ensemble models, and hybrid algorithms

These models are evaluated using:

• Accuracy

• Precision, Recall, F1-Score

• Confusion Matrix

• AUC-ROC Curve

4. Literature Review Highlights

• Traditional ML models (e.g., SVM, Naïve Bayes) perform well on short, simple reviews.

• Deep learning models (e.g., LSTM, CNN) excel with longer, complex text.

• LLMs like GPT-4 outperform traditional models in nuanced sentiment detection.

• Optimized methods (e.g., LSIBA-ENN, EGJO-LSTM) show high accuracy with custom feature engineering and weighting techniques.

• Some studies explore aspect-based sentiment analysis (ABSA) for more granular feedback.

• Explainable AI (XAI) is used to interpret model decisions and uncover biases.

5. Methodology

The proposed approach includes:

• A. Data Collection: Reviews are sourced and labeled.

• B. Preprocessing: Includes text cleaning, normalization, and balancing.

• C. Feature Extraction: Text is transformed into numerical form.

• D. Model Selection: Models are chosen based on task complexity.

• E. Training: Models are trained using techniques like cross-validation.

• F. Evaluation: Performance is measured using various metrics.

• G. Deployment: Models are integrated into applications via APIs.

• H. Continuous Improvement: Includes retraining, active learning, and drift detection.

6. Results and Insights

• Example results show sentiment distribution (e.g., 60% positive, 30% negative, 10% neutral).

• Confusion matrix and evaluation metrics confirm model effectiveness.

• LSTM may outperform others with up to 91% accuracy, especially for longer reviews.

• Insights include common issues (e.g., delivery delays) and product strengths (e.g., usability).

• Key performance example: Precision = 90%, Recall = 80%, F1-score = 85%, AUC = 0.9

7. Challenges

• Ambiguity and sarcasm in text complicate classification.

• Imbalanced datasets may bias model outcomes.

• Contextual understanding remains a limitation for some models.

Conclusion

In conclusion, sentiment analysis of customer product reviews using machine learning has become a valuable tool for businesses, providing automated insights into customer opinions. The methodology involves stages like data collection, preprocessing, model selection, and continuous improvement, enabling businesses to leverage feedback for better decision-making. Deep learning techniques have enhanced sentiment classification accuracy, allowing businesses to gain actionable insights from large datasets. Key benefits include improved customer insights, better decision-making, scalability, and real-time processing. Despite challenges such as ambiguity and domain-specific terminology, the need for continuous model improvement remains. Sentiment analysis empowers businesses to stay competitive, refine products, and enhance customer satisfaction in a dynamic market.

References

[1] J, I., & Kumar G, N., “Amazon Product Reviews Sentiment Analysis using Machine Learning”, JETIR, 234-239, 2022. [2] Ashbaugh, L., & Zhang, Y., “A Comparative Study of Sentiment Analysis on Customer Reviews Using Machine Learning and Deep Learning”, MDPI Computers, 1-16, 2024. [3] Daza, A., Gonzalez ´ Rueda, N., Sanchez, M. S., Espíritu, W. R., & Quinones, M. C., “Sentiment Analysis on E-Commerce Product Reviews Using Machine Learning and Deep Learning Algorithms: A Bibliometric Analysis,Systematic Literature Review, Challenges and Future Works”, International Journal of Information Management Data Insights, 1-20, 2024. [4] Ghatora, P. S., Hosseini, S. E., Pervez, S., Iqbal, M. J., & Shaukat, N., “Sentiment Analysis of Product Reviews Using Machine Learning. Big Data Cognitive Computing” MDPI, 1-18, 2024. [5] Pang, B., & Lee, L., “Opinion mining and sentiment analysis”, 2008. [6] Rasappan, P., Premkumar, M., Sinha, G., & Chandrasekara, K., “Transforming sentiment analysis for e-commerce product reviews: Hybrid deep learning model with an innovative term weighting and feature selection”, Information Processing & Management, 2024. [7] So, C., “What Emotions Make One or Five Stars? Understanding Ratings of Online Product Reviews by Sentiment Analysis and XAI”, 1st International Conference on Artificial Intelligence in HCI, AI-HCI. Kopenhagen, Denmark, Springer, 2020. [8] Taparia, A., & Bagla, T., “Sentiment Analysis: Predicting Product Reviews’ Ratings using Online Customer Reviews”, SSRN, 2020. [9] Xu, H., Liu, B., Shu, L., & Yu, P. S., “BERT Post-Training for Review Reading Comprehension and Aspect-based Sentiment Analysis”, arxiv logo, 2019. [10] Zhao, H., Liu, Z., Yao, X., & Yang, Q., “A machine learning-based sentiment analysis of online product reviews with a novel term weighting and feature selection approach”, Information Processing & Management, 2021.

Copyright

Copyright © 2025 Abhishek Kamble, Shruti Chandane, Anuja Kamble, Urmila Pol, Parashuram S. Vadar, Tejashree T. Moharekar. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.