Month 2 — Classical ML (Scikit-learn)

🎯 Goal of this month

By the end of the month you will be able to:

• Turn a real business problem into an ML task (regression / classification / clustering)
• Build a complete ML pipeline with Scikit-learn
• Evaluate a model and understand error types
• Create production-grade models with XGBoost/LightGBM
• Participate in a Kaggle competition and reach the top 30%

Weekly breakdown

Chapter order

1. Introduction to ML — terms, process, training/test split
2. Regression — predicting a continuous value
3. Classification — separating into classes
4. Clustering — unsupervised grouping
5. Feature Engineering — preparing and creating features
6. Model Evaluation — metrics and validation
7. Ensemble Methods — Random Forest, XGBoost, LightGBM
8. Exercises — additional exercises and Kaggle tasks

What can you do by the end of the month?

• Solve 80% of tabular data problems (regression, classification, clustering)
• Write reproducible code using Scikit-learn Pipeline
• Save a model with joblib and serve it via FastAPI
• Reach Kaggle top 30% with XGBoost/LightGBM
• Explain the ROI of an ML model to the business

Tip for Backend Dev

Like writing a REST API in backend, in ML there’s the fit() → predict() pattern:

# Backend pattern
@app.post("/users")
def create_user(data: UserIn):
    user = User(**data.dict())
    db.add(user)
    return user

# ML pattern
model = LogisticRegression()
model.fit(X_train, y_train)        # "train"
predictions = model.predict(X_test) # "predict"

This pattern is the same for all sklearn models — if you know LinearRegression, you also know RandomForest.

MLOps integration (from the start)

Backend dev’s advantage — production thinking. When writing your first model, already think about:

1. Reproducibility — random_state=42 everywhere
2. Saving — joblib.dump(model, 'model.pkl')
3. Versioning — model_v1.pkl, model_v2.pkl
4. Schema — input/output validation with Pydantic
5. Logging — timestamp and input for every prediction

These topics are covered deeper in Month 6 (MLOps), but start from day one.

Start

Start with Introduction to ML.