Advanced ML

Where Models Get Smarter

This section covers the techniques that power real production systems — ensemble methods, boosting, clustering, dimensionality reduction, and time series. These are the tools you reach for when a single model isn't enough or when the problem doesn't fit neatly into supervised learning.

AdaBoost

advanced

Each new model in the chain pays extra attention to whatever the last one got wrong.

Class Imbalance & SMOTE

intermediate

When 99% of your data is one class, your model can cheat by never predicting the other one.

Clustering Evaluation & Silhouette Score

intermediate

There are no labels to check against, so how do you know if your clusters are actually good?

Curse of Dimensionality

intermediate

The more features you add, the emptier and stranger your data's space becomes.

DBSCAN

advanced

Finds clusters of any shape by following dense neighborhoods — and knows when a point is just noise.

Ensemble Methods (Bagging vs Boosting)

intermediate

Why trust one model when you can average the opinions of a hundred?

Gradient Boosting

advanced

Instead of reweighting mistakes, each new model directly predicts the error of the last one.

Hierarchical Clustering

intermediate

Build a family tree of your data instead of committing to one fixed number of clusters upfront.

Hyperparameter Tuning

intermediate

The dials a model can't learn on its own — you have to search for the right settings.

K-Means Clustering

beginner

Group your data into k neat clusters — you just have to tell it how many to look for.

The Kernel Trick

advanced

Make non-linear data linearly separable — without ever actually computing the higher dimension.

Naive Bayes

beginner

A fast, surprisingly effective classifier built on one bold, 'naive' assumption.

Principal Component Analysis (PCA)

intermediate

Squeeze dozens of correlated features into a handful of ones that capture almost all the story.

Random Forest

intermediate

A forest of imperfect decision trees that together make surprisingly few mistakes.

Stacking & Blending

advanced

Train a model whose only job is to learn how to best combine your other models.

Support Vector Machines (SVM)

intermediate

Draw the widest possible street between classes, not just any line that separates them.

t-SNE

advanced

A dimensionality reduction technique built purely to make high-dimensional data visually intuitive.

XGBoost, LightGBM & CatBoost

advanced

The three production-grade gradient boosting libraries that actually win competitions.