arkodeep/spam-classfication-model

Spam Detection System

Lite Model

Introduction

The Lite model is a streamlined approach with optimized parameters and enhanced feature extraction designed for quick and efficient spam detection.

Features

• Text Preprocessing: Lemmatization, removal of stop words and punctuation.
• Feature Extraction: Text length, word count, unique word count, uppercase count, special character count.
• Model Creation: Ensemble model using SVC, MultinomialNB, and ExtraTreesClassifier.
• Visualization: Generates graphs for dataset insights, word clouds, and performance metrics.
• Metrics Saving: Accuracy, precision, and F1 score.

How to Run

1. Train the Model:

   python training/train_model_lite.py
   

2. Use the Model:

   import joblib
   model = joblib.load('models/model.pkl')
   vectorizer = joblib.load('models/vectorizer.pkl')
   

Legacy Model

Introduction

The Legacy model retains the original model logic without optimization but updates the structure and adds visualizations for spam detection.

Features

• Text Preprocessing: Porter Stemming, removal of stop words and punctuation.
• Model Creation: Ensemble model using SVC, MultinomialNB, and ExtraTreesClassifier with original parameters.
• Visualization: Generates graphs for dataset insights, word clouds, and performance metrics.
• Metrics Saving: Accuracy and precision.

How to Run

1. Train the Model:

   python training/train_model_legacy.py
   

2. Use the Model:

   import joblib
   model = joblib.load('models/model.pkl')
   vectorizer = joblib.load('models/vectorizer.pkl')
   

Additional Information

• Dependencies: Python 3.6 or higher, pip, and required packages listed in requirements.txt.
• Dataset: The dataset used for training is spam.csv.
• Contact and Support: For questions or support, please contact the project maintainers.

For more details, you can refer to the README.md and models.md.