A small machine,
trained to recognise
three irises.
Four measurements — sepal length, sepal width, petal length, petal width. Four classical algorithms, each trained on Fisher's 1936 dataset of 150 specimens. Adjust the dials, watch every model agree (or quietly disagree) on the species.
“The use of multiple measurements in taxonomic problems.”
— R. A. Fisher
Measure a specimen.
Adjust the dials below — every value is in centimetres. The four trained models will each cast a vote.
150 flowers, plotted.
Each dot is a real iris from Fisher's herbarium. Choose any two measurements to see how the species cluster — or overlap.
Plot any two features
- Iris setosa
- Iris versicolor
- Iris virginica
Sepal Length
Sepal Width
Petal Length
Petal Width
Four models. One verdict.
Each algorithm was trained on a 75/25 split of the dataset and evaluated on the held-out 25%. Metrics are macro-averaged across the three species.
Accuracy comparison
Logistic Regression
92.11%A linear probabilistic classifier — the lab notebook baseline.
K-Nearest Neighbors
92.11%Classifies by majority vote of the 5 closest specimens.
Support Vector Machine
94.74%Finds the maximum-margin hyperplane (RBF kernel).
Decision Tree
92.11%Recursive feature splits — interpretable like a field key.