Skip to learning content

DATA SCIENCE PYTHON PLAYGROUND

Machine Learning · Learn / Refresh

← Choose a deckQUESTIONS · MODELS · EVIDENCE
ROUTE 3B

Classification

Understand class evidence and nine model families.

01 / Evidence

C01Core · teaching
Read a confusion matrix: True class, Predicted class, Count each errorTPFPFNTNTrue classPredicted classCount each error

Read a confusion matrix

Track which labels are confused.

12–18 min · 3 practices
C02Core · teaching
Precision and recall: Precision, Recall, Error costsTPFPFNTNPrecisionRecallError costs

Precision and recall

Connect different errors to different questions.

12–18 min · 3 practices
C03Core · teaching
Macro F1 and imbalance: Class A F1, Class B F1, Macro averageClass A F1Class B F1Macro average

Macro F1 and imbalance

Evaluate every class explicitly.

12–18 min · 3 practices

02 / Probabilities and neighbours

C04Core · teaching
Labels and probabilities: Class chance, Threshold, Final labelClass chanceThresholdFinal label

Labels and probabilities

Align probability columns with class labels.

12–18 min · 3 practices
C05Core · teaching
Logistic regression: Linear boundary, Class A, Class BLinear boundaryClass AClass B

Logistic regression

Understand a regularised linear class boundary.

12–18 min · 3 practices
C06Core · teaching
A logistic workflow: Scale inside CV, Logistic fit, Macro F1VTTTTVTTTTVTTTTVScale inside CVLogistic fitMacro F1

A logistic workflow

Tune regularisation and investigate shortcuts using training evidence.

12–18 min · 3 practices
C-R1Core · review
Classification evidence retrieval: Confusion counts, Macro F1, Reference scoreConfusion countsMacro F1Reference score

Classification evidence retrieval

Retrieve earlier concepts before combining them.

15–20 min · 3 practices
C07Go Further · teaching
Threshold choices depend on costs: Lower threshold, More alerts, Fewer missesLower thresholdMore alertsFewer misses

Threshold choices depend on costs

Choose thresholds using training-only evidence and explicit requirements.

18–25 min · 4 practices
C08Core · teaching
Neighbour voting: Query point, Nearby votes, Chosen labelQuery pointNearby votesChosen label

Neighbour voting

Explain a classifier’s local distance-based decision.

12–18 min · 3 practices
C09Core · teaching
Validate k on a meaningful scale: Small k, Large k, Validate kSmall kLarge kValidate k

Validate k on a meaningful scale

Choose k with fold-local scaling.

12–18 min · 3 practices

03 / Margins and rules

C10Core · teaching
Margins and support vectors: Margin, Support rows, Scaled SVCMarginSupport rowsScaled SVC

Margins and support vectors

Interpret the separating boundary and supporting examples.

25–40 min · 4 practices
C11Core · teaching
Nonlinear SVM behaviour: RBF boundary, Curved regions, Tune with CVRBF boundaryCurved regionsTune with CV

Nonlinear SVM behaviour

Relate C, kernel scale and cost to validation.

12–18 min · 3 practices
C12Core · teaching
One feature, one rule: One feature, Majority rule, Default classOne featureMajority ruleDefault class

One feature, one rule

Understand the production One-R rule and fallback.

12–18 min · 3 practices
C13Core · teaching
One-R with numeric inputs: Numeric cut, Binned values, One-R ruleNumeric cutBinned valuesOne-R rule

One-R with numeric inputs

Keep numeric bin learning inside each fit.

12–18 min · 3 practices
C-R2Core · review
Neighbours, margins and rules retrieval: Neighbour scale, SVM margin, One-R ruleNeighbour scaleSVM marginOne-R rule

Neighbours, margins and rules retrieval

Retrieve earlier concepts before combining them.

15–20 min · 3 practices
C14Core · teaching
Classification trees: Class counts, Purer leaves, Predicted classClass countsPurer leavesPredicted class

Classification trees

Learn recursive splits, impurity, class leaves and validation within Classification.

18–25 min · 4 practices
C-K1Core · checkpoint
Classification checkpoint: Penguin X, Choose by CV, Inspect confusionPenguin XChoose by CVInspectconfusion

Classification checkpoint

Complete a stratified mixed Penguin logistic workflow with macro F1, a reference, tuning and confusion evidence.

25–35 min · 1 practices

04 / Class distributions

C15Core · teaching
Gaussian Naive Bayes: Class prior, Feature density, PosteriorClass AClass BClass priorFeature densityPosterior

Gaussian Naive Bayes

Combine class priors with continuous feature evidence.

25–40 min · 4 practices
C16Core · teaching
Match Naive Bayes to feature types: Binary counts, Word counts, Continuous XBinary countsWord countsContinuous Xbinary0 / 1Bernoullimeasuredreal valueGaussian

Match Naive Bayes to feature types

Use each production Naive Bayes preparation path.

18–25 min · 4 practices
C17Core · teaching
LDA shares a covariance shape: Different means, Shared shape, LDADifferent meansShared shapeLDA

LDA shares a covariance shape

Understand shared class geometry and linear boundaries.

12–18 min · 3 practices
C18Core · teaching
QDA allows class-specific shapes: Class A shape, Class B shape, QDAClass A shapeClass B shapeQDA

QDA allows class-specific shapes

Balance covariance flexibility against data support and stability.

18–25 min · 4 practices
C-R3Core · review
Trees and distributions retrieval: Tree leaves, Bayes likelihood, LDA or QDATree leavesBayes likelihoodLDA or QDA

Trees and distributions retrieval

Retrieve earlier concepts before combining them.

15–20 min · 3 practices
C19Go Further · teaching
Covariance support and regularisation limits: Few rows, Covariance fit, RegulariseFew rowsCovariance fitRegularise

Covariance support and regularisation limits

Qualify numerical stability and unit effects.

12–18 min · 3 practices