How tolgatopaloglu/turkish-sms-spam-detection works
This repository provides Turkish SMS spam classification: main.py loads a serialized model, predicts labels for sample messages, prints the results, and writes test.csv. The README also documents a separate model-development workflow using TF-IDF, model comparison and tuning, metric tracking, and model registration; the notebook and dataset are present, but their implementation was not sampled, so their internal wiring is shown only where documented. The trained artifact's internal pipeline is described by the README, not verified from source.
Main entry points: turkish_SMS_spam_classification.ipynb Training workflow · main.py Prediction script · SMS_Spam_Classifcation.pkl Trained model · test.csv CSV results
Components 16
SMS inference
- Prediction script
main.py - Sample messages
main.py - Trained model
SMS_Spam_Classifcation.pkl - TF-IDF features
SMS_Spam_Classifcation.pkl - Tuned LinearSVC
SMS_Spam_Classifcation.pkl - Spam / Normal
main.py - Result table
main.py
Model development
- Training workflow
turkish_SMS_spam_classification.ipynb - Model comparison
turkish_SMS_spam_classification.ipynb - Hyperparameter tuning
turkish_SMS_spam_classification.ipynb - Evaluation metrics
turkish_SMS_spam_classification.ipynb - Model registry
turkish_SMS_spam_classification.ipynb
Data and artifacts
- CSV results
test.csv - SMS dataset
TurkishSMSCollection.csv
Other components
- Project runner
- MLflow
Connections 17
- Project runner → Prediction script: runs
- Prediction script → Trained model: loads
- Prediction script → Sample messages: predicts on
- Prediction script → Trained model: calls predict
- Trained model → TF-IDF features: transforms with
- TF-IDF features → Tuned LinearSVC: provides features
- Tuned LinearSVC → Spam / Normal: classifies as
- Prediction script → Result table: builds
- Spam / Normal → Result table: adds predictions
- Result table → CSV results: writes
- Training workflow → SMS dataset: uses
- Training workflow → Model comparison: compares
- Training workflow → Hyperparameter tuning: optimizes
- Training workflow → Evaluation metrics: tracks
- Training workflow → MLflow: logs experiments
- Training workflow → Model registry: registers model
- Training workflow → MLflow: uses registry