patrick-paul/ssd-training

Swahili Spam Detection (SSD) - Training Model # Swahili Spam Detection - Training Repository The training component of the Swahili Spam Detection system, containing datasets, training notebooks, and model evaluation infrastructure. This repository handles the machine learning aspects of the main SSD project. ## Project Structure ```bash └── ssd-training/ └── dataset/ # Training datasets └── combined_set.csv └── combined_set.xlsx └── Model_After_Training/ └── model_after_tranining_28_jan_2025/ └── swahiliSpamDetectionModel.pkl └── swahiliSpamDetectionModel.pkl └── stopwords/ # Swahili stopwords └── Common Swahili Stop-words.csv └── spamDetectionRef.ipynb # Training notebook └── requirements.txt # Dependencies ``` ## Model Performance ### Performance Metrics | Model | Technique | Accuracy | Precision | Recall | F1-score | AUC-ROC | |---------------------|---------------------|----------|-----------|--------|----------|---------| | Logistic Regression | Count Vectorization | 0.9924 | 0.9925 | 0.9924 | 0.9923 | 0.9996 | | Naive Bayes | Count Vectorization | 0.9904 | 0.9905 | 0.9904 | 0.9905 | 0.9988 | | SVM | Count Vectorization | 0.9933 | 0.9934 | 0.9933 | 0.9933 | 0.9995 | | Random Forest | Count Vectorization | 0.9933 | 0.9933 | 0.9933 | 0.9933 | 0.9993 | | Logistic Regression | TF-IDF | 0.9838 | 0.9842 | 0.9838 | 0.9837 | 0.9995 | | Naive Bayes | TF-IDF | 0.9952 | 0.9952 | 0.9952 | 0.9952 | 0.9985 | | SVM | TF-IDF | 0.9914 | 0.9915 | 0.9914 | 0.9914 | 0.9999 | | Random Forest | TF-IDF | 0.9924 | 0.9925 | 0.9924 | 0.9923 | 0.9998 | ## Getting Started ### Prerequisites - Python 3.8+ - pip package manager ### Installation 1. Clone the repository ```bash git clone github.com cd ssd-training ``` 2. Create virtual environment ```bash python …