Movie Review Sentiment Analysis

Project Overview

This project aims to classify movie reviews as positive or negative based on their textual content. It follows a clean, production-style machine learning pipeline with proper project structure.

Dataset

Source: IMDB Movie Reviews Dataset

Size: 50,000 reviews

Classes: Positive / Negative (balanced)

Dataset is located in: /data

Project Structure

Sentiment_Classifier/ │

├── data/

│ └── IMDB_Dataset.csv

│

├── notebooks/

│ ├── 01_eda_and_cleaning.ipynb

│ ├── 02_feature_engineering_and_training.ipynb

│ └── 03_model_evaluation.ipynb

│

├── models/

│ ├── Sentiment_Classifier_v1/

│ └── Sentiment_Classifier_v2/

│

├── experiments/

│ └── test_models.ipynb

│

├── utils/

│ └── preprocessing.py

│

│

├── README.md

Methodology

1. Exploratory Data Analysis

Sentiment distribution

Review length analysis

2. Text Preprocessing

HTML tag removal

Lowercasing

Punctuation & number removal

Stopword removal

Lemmatization

3. Feature Engineering

TF-IDF Vectorization

4. Model Training

Logistic Regression (baseline)

Support Vector Machine (experiments)

5. Evaluation

Precision, Recall, F1-score

Confusion Matrix

Results

The dataset is perfectly balanced (50% positive / 50% negative)

Baseline Logistic Regression achieved strong F1-score performance

Detailed evaluation is available in 03_model_evaluation.ipynb