Amazon Product Review Classification

• This project was focused on the development of a machine learning-based classifier designed to accurately categorize Amazon product reviews.
• By leveraging advanced algorithms such as Naive Bayes, SVM, Logistic Regression, XGBoost, and Random Forest, the classifier is able to achieve 89% accuracy and an 87% F1-score.
• In this project, I have used TF-IDF vectorizer for feature extraction to interpret and classify vast quantities of text data effectively.

---

Project Overview

• Objective: To accurately classify Amazon product reviews using machine learning algorithms.
• Algorithms Used: Naive Bayes, SVM, Logistic Regression, XGBoost, and Random Forest.
• Feature Extraction: Utilized TF-IDF.
• Validation Method: 5-fold cross-validation.
• Dataset: 220,909 Amazon product reviews (Link).

Technologies Used

• Python: The primary programming language used for model development.
• Jupyter Notebook: For documenting the development process and performing data analysis.
• Scikit-learn: Utilized for model training, feature extraction, and evaluation.
• XGBoost: For implementing the XGBoost algorithm.
• Pandas & NumPy: For data manipulation and numerical calculations.
• NLTK: For text preprocessing and natural language processing tasks such as tokenization, stemming, lemmatization and parsing.

Algorithm Performance

Algorithm	Accuracy (%)	F1-Score (%)
Naive Bayes	88	83
Random Forest	88	82
XGBoost	88	85
Logistic Regression	89	86
SVM	89	87