Supervised Learning in Depth

Course outline and learning expectations

This is a tutor-led course. The outline shows what your tutor will cover; teaching materials and examinations are provided directly to enrolled students.

What you will learn

• Build and evaluate multiple linear regression models
• Train decision tree, random forest, k-Nearest Neighbours and support vector machine models
• Prepare categorical variables using suitable encoding methods
• Apply train-test splits, k-fold and stratified cross-validation
• Tune model hyperparameters using grid and randomised search basics
• Compare models using metrics suited to regression and classification problems
• Identify data leakage, overfitting and important model limitations
• Complete and communicate a supervised learning capstone using weather and soil data

Modules

1. 1

   Regression deep dive

   Build and evaluate multiple linear regression models for continuous outcomes.

   Multiple linear regressionFeature selection basicsRegression assumptionsResidual analysisMAEMSERMSER-squared
2. 2

   Classification with decision trees and random forests

   Train and assess tree-based classifiers for structured data.

   Decision treesRandom forestsFeature importanceClass imbalance basicsAccuracyPrecisionRecallF1 scoreROC-AUC
3. 3

   k-Nearest Neighbours and support vector machines

   Apply distance-based and margin-based classification methods.

   Feature scalingChoosing kDistance metricsSupport vector machinesKernel basicsDecision boundaries
4. 4

   Categorical data and encoding

   Prepare categorical variables consistently for machine learning pipelines.

   Nominal and ordinal dataOne-hot encodingOrdinal encodingColumnTransformerPipelinesPreventing data leakage
5. 5

   Cross-validation

   Estimate model performance more reliably across different data splits.

   Holdout validationk-fold cross-validationStratified k-foldCross-validation scoresBias and variance
6. 6

   Hyperparameter tuning basics

   Improve model settings through structured search and validated scoring.

   Model parameters and hyperparametersGridSearchCVRandomizedSearchCVScoring metricsInterpreting tuning results
7. 7

   Comparing multiple models

   Evaluate several algorithms on the same problem using a consistent process.

   Baseline modelsShared preprocessingMetric selectionError analysisInterpretabilityModel selection
8. 8

   Capstone supervised learning project

   Predict crop disease risk from weather and soil data using three algorithms and compare results.

   Problem framingData cleaningFeature engineeringTraining three modelsCross-validation and tuningModel comparisonLimitationsProject presentation