fix: replace wget with python urllib to download models — wget not in slim image

fix: download models from GitHub Releases instead of HF hub

Upload models/scaler.pkl with huggingface_hub

Upload models/xgboost_tuned.pkl with huggingface_hub

fix: add huggingface_hub to requirements so start.sh can import it

hf spaces: add space metadata to README

hf spaces: download models at startup instead of COPY

hf spaces: port 7860

gitignore updated

fix: accept HEAD requests on /health so UptimeRobot monitoring works

read me updated

Trigger CI rum

Step 10 + 11: add CI/CD pipeline and Render deployment

Step 9: add Docker containerization

docs: expand MLflow section with UI walkthrough and errors; add interview pipeline guide

Step 8: add 28 automated tests covering preprocessing, models, and API

Step 7: FastAPI with /predict /health endpoints, V1-V28 inference explanation in README

Step 6: select xgboost_tuned as final model with full comparison and reasoning in README

Step 6: switch to CatBoost as final model, add tune_catboost.py, rename xgboost_final to xgboost_tuned

README: rewrite with generated files explained inside each step in plain language

Step 6: model comparison charts, XGBoost GridSearchCV tuning, real benchmark scores in README

Step 5: remove SVM/KNN/MLP, clarify 10 pkl files in models dir (9 models + scaler)

README: explain every created file under each step section

README: complete rewrite with all steps, actual file tree, and full details for Steps 1-4

Step 4: MLflow setup with SQLite backend and experiment tracking helpers

Step 3: add generated file details to README

Step 3: data preprocessing pipeline with scaling, train/test split, and SMOTE

Step 2: add jupyter/nbconvert to requirements and EDA section to README

Step 2: EDA notebook with class imbalance, amount, time and feature analysis

Step 1: project setup, folder structure, and environment