Getting started

1) Install dependencies

make requirements

2) Prepare a dataset scenario

Choose a scenario from data/raw (valve1, valve2, other, anomaly-free, or all):

make dataset DATA_SCENARIO=valve2
make features DATA_SCENARIO=valve2

Tip: avoid all if any source files are missing required label columns.

3) Run training and inference experiments

make train MODEL=isolation_forest DATA_SCENARIO=valve1 
make predict MODEL=isolation_forest DATA_SCENARIO=valve1 

Optional MLflow UI:

make mlflow_ui

4) Start Grafana + Prometheus services

make monitoring_up

Service URLs:

• Grafana: http://localhost:3000 (admin / admin)
• Prometheus: http://localhost:9090
• Pushgateway: http://localhost:9091

In Grafana, inspect:

• Recent Train and Predict Runs
• Operational Health
• MLflow Experiment Quality

Next reading

• Dataset details: Dataset (SKAB)
• Model behavior: Models
• Tracking and registry: MLflow
• Dashboards and metric paths: Monitoring

5) Stop monitoring services

make monitoring_down