cs179-imdb

This project is an exploration-driven research study on improving the Multinomial Naive Bayes (MNB) model for sentiment classification. Using the IMDb movie review dataset from Kaggle, we investigate three distinct methods to enhance the model’s performance — focusing on accuracy, feature quality, and computational efficiency — while preserving the model’s simplicity and interpretability.

Research Objective

While MNB is widely used for text classification due to its simplicity and interpretability, it has known limitations. This project explores three complementary methods to improve MNB:

1. Hyperparameter Tuning
   Fine-tuning model parameters (e.g., smoothing factor alpha, n-gram range) to increase predictive accuracy.

2. Enhancement Techniques
   Improving feature representation by:

   • Replacing CountVectorizer with TfidfVectorizer
   • Boosting the importance of discriminative keywords identified from the model

3. Efficiency Optimization (Planned)
   Applying methods such as dimensionality reduction and sparse matrix optimization to reduce runtime and memory usage without sacrificing performance.

---

📘 Systematic Improvement of Multinomial Naive Bayes Classifier for Sentiment Classification: Parameter Tuning, Enhancement Techniques, and Efficiency Optimization

Project Structure

.
├── data
│   └── IMDB-Dataset.csv
├── efficiency_optimization
│   ├── DiminishingLatency_ConfusionMatrix.png
│   └── MNB_Model_Diminishing_Latency_Study.py
├── enhancement_techniques
│   └── MNB_Model_Enhancement_Techniques_Study.ipynb
├── MNB_base_model.py
├── parameter_tuning
│   └── MNB_Model_ParameterTuning.py
└── README.md