Wine Cultivar Classification
Compare k-NN, decision tree, and random forest classifiers using 13 chemical measurements from the UCI Wine dataset.
Try a sample
Predictions run in your browser. Your inputs are not sent to a prediction service.
The output is a cultivar class, not a wine quality rating or a named grape variety. Probabilities are uncalibrated model estimates.
Model comparison
142 training samples and 36 test samples. Scaling and hyperparameter tuning use training data only, with five-fold stratified cross-validation. The default classifier is selected by training CV.
| Classifier | Training CV accuracy | Test accuracy | Test macro F1 |
|---|
Silhouette analysis
K-means cluster separation on the training partition. Silhouette is not classification accuracy.
| Clusters | Silhouette score |
|---|
These results are from a small public benchmark. They do not establish performance on new wine sources.