IMDB Sentiment Analysis Model

📌 Project Overview

This repository contains a sentiment analysis project trained on the IMDb 50,000-review dataset (25k positive, 25k negative). The model classifies whether a movie review is positive or negative using deep learning (Keras/TensorFlow). A saved tokenizer (tokenizer.pkl) and trained model (IMDB_model.h5) are used for inference.

📁 Dataset

• Dataset: IMDb Movie Review Dataset (50,000 reviews)
• Public sources:
  • Kaggle: https://www.kaggle.com/datasets/lakshmi25npathi/imdb-dataset-of-50k-movie-reviews
  • Stanford: http://ai.stanford.edu/~amaas/data/sentiment/
• Format: CSV containing review and sentiment columns

✅ Example row from dataset:

"One of the other reviewers has mentioned that after watching just 1 Oz episode you'll be hooked...",positive

🧠 Project Structure

project-root/
├── data/                     # raw dataset (ignored in Git)
├── models/                   # saved model + tokenizer (ignored)
│   ├── IMDB_model.h5
│   └── tokenizer.pkl
├── src/                      # preprocessing + training scripts
├── notebooks/                # optional Jupyter notebooks
├── README.md
├── requirements.txt
└── .gitignore

⚙️ Model Pipeline

1. Load and preprocess text (clean HTML, lowercase, remove punctuation, etc.)
2. Tokenize and pad sequences (tokenizer.pkl)
3. Train a neural network (LSTM/CNN/etc.) on padded sequences
4. Save trained model (IMDB_model.h5) and tokenizer for later inference
5. Load model + tokenizer to classify new reviews

❗ Why Model & Dataset Files Aren’t in Git

Large binary artifacts (datasets, .h5 models, .pkl tokenizers) are not committed to Git because:

• They quickly bloat repo size
• GitHub hard-limits single files >100MB
• They update frequently and do not diff well

Instead, they should be stored using:

• Git LFS
• HuggingFace Hub
• Google Drive / S3 / GitHub Releases
• Or a download script (download_data.py)

📋 Requirements

• Python 3.x
• numpy
• pandas
• tensorflow / keras
• scikit-learn
• nltk (or similar for text cleaning)

Install using:

pip install -r requirements.txt

▶️ Usage

1. Clone repository

git clone https://github.com/<your-username>/IMDB-Sentiment-Analysis-Model.git
cd IMDB-Sentiment-Analysis-Model

2. Create virtual environment

macOS/Linux:

python -m venv .venv
source .venv/bin/activate

Windows:

python -m venv .venv
.venv\Scripts\activate

3. Install dependencies

pip install -r requirements.txt

4. Download dataset

• Place CSV file into data/ OR use a download script if provided.

5. Train or run inference

python src/train.py      # train model
python src/predict.py    # classify new text

✅ Evaluation

Add accuracy, loss curves, confusion matrix, or sample predictions here.

📜 Acknowledgements

• Dataset: “Large Movie Review Dataset” — Andrew Maas et al. (Stanford AI Lab, 2011)
• Code: Created by
• Libraries: TensorFlow, Keras, NumPy, Pandas, Scikit-Learn

🤝 Contributing

Pull requests are welcome — feel free to open issues or suggest improvements.

⭐ If you found this useful, consider giving the repo a star!