Stage 1 · Understand · Beginner
Machine Learning Basics
Learn how machine learning models are built, tested and judged. Frame a problem, prepare data, train classic models with scikit-learn, evaluate them honestly, avoid overfitting and leakage, and know when to use clustering, anomaly detection, a rule or a language model.
7 lessons · 12 hr · Free
You will learn to
- ▸Frame a business problem as classification, regression, clustering or anomaly detection
- ▸Prepare data, split it properly and spot data leakage
- ▸Train and compare linear, tree-based and nearest-neighbor models against a baseline
- ▸Evaluate with confusion matrices, precision, recall, thresholds, costs and cross-validation
- ▸Recognize and reduce overfitting, and tune settings without peeking at the test set
- ▸Choose between machine learning, simple rules and language models, and monitor a model for drift
Before you start
- ·Comfortable with Python and pandas, as taught in Python and Data Essentials
- ·No advanced maths is required, only comfort with percentages and averages
- ·Recommended: AI Foundations for IT Professionals first
Syllabus
- 01Machine learning in practiceKinds of learning, the workflow, and when machine learning is the right tool.1 hr
- 02Data, features and honest splitsFeatures and labels, train and test sets, leakage and cleaning.1 hr 30 min
- 03Your first modelsBaselines, regression, logistic regression, trees and nearest neighbors.2 hr
- 04Evaluating modelsThe confusion matrix, precision, recall, thresholds, costs and cross-validation.1 hr 30 min
- 05Overfitting and improvementSpotting overfitting, regularization, tuning and ensembles.1 hr 30 min
- 06Clustering, anomalies and choosing toolsUnsupervised learning, drift, responsible use and rules versus models versus LLMs.1 hr 30 min
- 07Capstone: a server failure early warningBuild, evaluate and write up a complete model, with its limits.3 hr