Machine Learning
Where Data Starts Learning
Machine learning is what happens when you let data find the patterns instead of programming them by hand. This section covers how models learn, the core algorithms, how to evaluate whether a model is actually good, and the fundamental tradeoffs every ML practitioner deals with. It's the heart of modern data science, and understanding it is key to using it effectively.
What is Machine Learning?
beginnerTeaching computers to learn from examples instead of following a fixed set ofrules.
Confusion Matrix
beginnerA simple 2x2 table that shows exactly where your classifier is getting confused.
Precision and Recall
beginnerPrecision: of the ones you said yes to, how many were right? Recall: of the actual yeses, how many did you catch?
ROC Curve and AUC
intermediateA tool to see how well your classifier separates the two classes at every possible threshold.
Decision Trees
intermediateFlowchart-style questions that split your data into smaller groups until you can make a prediction.
K-Nearest Neighbors (KNN)
intermediateTell me who your neighbors are, and I will tell you what you are.
Feature Scaling
beginnerPutting all your features on the same ruler so no single one dominates just because it has bigger numbers.
Handling Missing Data
beginnerReal data has holes. You have to decide what to fill them with – or whether to leave them be.
Training vs Inference
beginnerLearning is one phase. Using what you learned is another. Dont mix them up.
Underfitting vs Overfitting
beginnerToo simple and you miss the pattern. Too complex and you memorize the noise. Goldilocks is in the middle.
Bias-Variance Tradeoff
intermediateEvery model has two sources of error: being too rigid (bias) or being too jumpy (variance). You cannot eliminate both.
MAE vs MSE vs RMSE vs R²
beginnerFour ways to measure how wrong your model is — and why it matters which one you pick.
Train-Test Split
beginnerNever test your model on the data it trained on. That is like giving students the exam questions during lectures.
Cross-Validation
intermediateWhen you cannot afford to waste data on a single test set, use this trick to get a honest performance estimate.
Linear Regression
beginnerThe straight line through your data. Simple, interpretable, and the foundation of almost everything else.
Regularization
intermediateA trick to stop your model from overfitting by punishing it for being too complex.
Logistic Regression
intermediateA linear model for yes/no questions. It gives you probabilities, not just guesses.