How dimus0/telco-churn-dvc-mlflow works
This repository is a Telco customer-churn machine-learning workflow. Raw data is encoded, balanced, and split into prepared CSVs; users can train a configured random forest directly or run Optuna hyperparameter search, then record metrics and model artifacts with MLflow. The source excerpts establish these script-level data flows, but the README provides no additional capability descriptions. DVC files and configuration files are present, though their runtime wiring is not shown in the excerpts.
Main entry points: optimize.py Optimize parameters · train.py Train classifier · prepare.py Encode and split · prepared Prepared splits
Components 12
Data preparation
- Raw churn data
data/raw/telco_dataset.csv.dvc - Encode and split
src/prepare.py - Prepared splits
data/prepared
Modeling
- Train classifier
src/train.py - Model settings
config/model/random_forest.yaml
Hyperparameter search
- Optimize parameters
src/optimize.py - Workflow settings
config
Tracking and outputs
- MLflow tracking
- Saved model
- Saved metrics
- Run artifacts
mlartifacts
Other components
- Workflow user
Connections 16
- Raw churn data → Encode and split: reads
src/prepare.py - Encode and split → Prepared splits: writes splits
src/prepare.py - Workflow user → Encode and split: runs
- Prepared splits → Train classifier: loads splits
src/train.py - Model settings → Train classifier: sets defaults
- Train classifier → MLflow tracking: logs run
src/train.py - Train classifier → Saved model: saves model
src/train.py - Train classifier → Saved metrics: writes metrics
src/train.py - MLflow tracking → Run artifacts: stores artifacts
- Workflow user → Train classifier: runs
- Workflow settings → Optimize parameters: configures
src/optimize.py - Optimize parameters → Prepared splits: loads data
src/optimize.py - Optimize parameters → MLflow tracking: logs trials
src/optimize.py - Optimize parameters → Saved model: saves model
src/optimize.py - Optimize parameters → Saved metrics: writes metrics
src/optimize.py - Workflow user → Optimize parameters: runs