← Back to Dashboard
Module 02 · Days 8–14
Classical Machine Learning

The algorithms running 80% of production ML today. You will understand them deeply — not just sklearn.fit() — so you can debug, tune, and explain them to any stakeholder.

D8
Supervised learning: regression from scratch
Linear regression, cost functions, gradient descent. Write it manually once; debug it confidently forever.
D9
Classification: logistic regression & decision boundaries
From predicting numbers to predicting classes. Sigmoid, cross-entropy, and what a decision boundary really is.
D10
Tree-based models: Decision Trees → XGBoost
How trees work, why Random Forests fix them, and why XGBoost wins tabular competitions.
D11
Model evaluation: the metrics that actually matter
Accuracy is a trap. Learn precision, recall, F1, AUC-ROC, and when each is the right measure.
D12
Unsupervised learning: clustering & PCA
K-Means, DBSCAN, and dimensionality reduction. Finding structure when you have no labels.
D13
Feature engineering & sklearn Pipelines
Good features beat better algorithms. Build Pipelines that are safe from data leakage and ready for production.
🏗
🏗️ PROJECT: End-to-End ML Classifier Pipeline
Real tabular dataset (Telco Churn, Credit Risk, or your own). EDA → feature engineering → 3 models → report.