Sentiment Analysis of Amazon Office Product Reviews

Project Overview This project performs sentiment analysis on Amazon office product reviews using various machine learning techniques. The goal is to classify reviews as positive or negative based on their content.

Table of Contents Dataset Preparation Data Cleaning Pre-processing TF-IDF Feature Extraction Machine Learning Models Results Requirements Usage

Dataset Preparation -The dataset is sourced from Amazon reviews for office products. -We use pandas to read the TSV file from a URL and load it into a DataFrame. -Only the 'review_body' and 'star_rating' columns are retained for analysis.

Data Cleaning The following cleaning steps are performed: Convert text to lowercase Remove HTML tags and URLs Expand contractions Remove non-alphabetic characters and special symbols Remove extra spaces

Pre-processing Text pre-processing includes: Removing stop words Lemmatization to reduce words to their base form TF-IDF Feature Extraction TF-IDF (Term Frequency-Inverse Document Frequency) is used to convert text data into numerical features. 8000 features are extracted using scikit-learn's TfidfVectorizer.

Machine Learning Models Four different models are implemented and compared: Perceptron Support Vector Machine (SVM) Logistic Regression Naive Bayes

Results

Performance metrics (Accuracy, Precision, Recall, F1 Score) are calculated for each model on both training and testing datasets.

Requirements Python 3.x pandas numpy nltk scikit-learn BeautifulSoup matplotlib

Clone the repository: git clone https://github.com/sriramgurazada/Sentimental-Analysis.git

Run the script: python sentimentalAnalysis.py