Every module includes three substantial lessons, downloadable source files, practical exercises, explained solutions and a knowledge check. Modules marked free are available under the current public access rules. Protected self-learning modules are free after registration and email verification. Publication, prerequisites and explicit private-access rules still apply.
01Free module
What machine learning can—and cannot—do
Frame a prediction problem and decide whether a model adds useful evidence.
- A prediction contract and timing diagram
- A learning-task classification
- A suitability and risk decision note
Overview & practice files ↗View lesson outline
- Prediction versus explanation
- Supervised and unsupervised learning
- Deciding whether ML is appropriate
02Free module
Your first prediction experiment
Use a separate pilot dataset to compare a fixed model with a train-only baseline.
- A feature/target timing audit
- A fixed pilot training and test split
- A baseline comparison and error note
Overview & practice files ↗View lesson outline
- Features and targets
- Training and test data
- Compare a simple model with a baseline
03Free · verified account
Preparing data for machine learning
Fit transformations within the training partition and prevent direct and indirect leakage.
- A train-only preprocessing record
- An encoding and missingness test log
- A leakage audit with explicit excluded features
View access requirements ↗View lesson outline
- Encoding categories · Verified account
- Missing values and scaling · Verified account
- Preventing data leakage · Verified account
04Free · verified account
Regression and classification
Inspect transparent linear and tree models and compare their errors on validation data.
- A linear regression error review
- A depth-one decision-tree explanation
- A metric and error-cost comparison
View access requirements ↗View lesson outline
- Linear models and decision trees · Verified account
- Choosing evaluation measures · Verified account
- Interpreting prediction errors · Verified account
05Free · verified account
Improving and evaluating models responsibly
Use time-aware validation, freeze decisions and report uncertainty and limitations honestly.
- A rolling validation and threshold record
- A final held-out evaluation and subgroup review
- A model card with a go/pause decision
View access requirements ↗View lesson outline
- Cross-validation for future prediction · Verified account
- Tuning, thresholds and overfitting · Verified account
- Imbalance, fairness and uncertainty · Verified account
06Free · verified account
Unsupervised Learning and ML Capstone
Describe clusters carefully and integrate a leakage-aware final evidence review.
- A similarity and feature-timing contract
- A tested cluster/count and sensitivity report
- An integrated supervised/unsupervised capstone handover
View access requirements ↗View lesson outline
- Frame unsupervised similarity without inventing labels · Verified account
- Implement and test a small k-means workflow · Verified account
- Integrate the ML evidence and hand over the capstone · Verified account
07Free · verified account
Demand forecasting and maintenance-risk evaluation
Separate forecast selection from future evaluation: use an original fictional source to implement, verify and explain three connected professional assignments.
- Source/metric/access contract
- Independent implementation and exact run instructions
- Unchanged control and changed-input/failure evidence
- Decision memo with limitations and next evidence
View access requirements ↗View lesson outline
- Separate forecast selection from future evaluation · Verified account
- Evaluate a predictive-maintenance rule without leaking machines · Verified account
- Write an uncertainty and monitoring handover · Verified account
