Simple ML Model

Two machine learning scripts exploring different levels of model complexity — from a minimal regression baseline to a neural network with model monitoring.

Scripts

• main-sklearn.py — Linear regression with scikit-learn on synthetic data
• main-2.py — MLP classifier with PyTorch on the Titanic dataset, including Evidently AI monitoring and Prometheus metrics

main-2.py — Titanic Survival Prediction

Dataset

The Titanic dataset is loaded automatically from a public URL. No local file needed.

Features used: Pclass, Sex, Age, SibSp, Parch, Fare, Embarked
Target: Survived (0 = No, 1 = Yes)

Model Architecture

A 3-layer MLP built with PyTorch:

• Loss: CrossEntropyLoss
• Optimizer: SGD (lr=0.01)
• Epochs: 1000
• Split: 80% train / 20% test

Monitoring

After training, the script generates:

• model_monitoring_report.html — Evidently AI report (Data Drift, Target Drift, Classification metrics)
• Prometheus metrics exposed at http://localhost:8000

How to Run

pip install torch pandas scikit-learn evidently prometheus-client

# Minimal linear regression
python main-sklearn.py

# Titanic MLP + monitoring
python main-2.py

Tech Stack

Python · PyTorch · scikit-learn · pandas · Evidently AI · Prometheus