Classification Algorithms Arena

Classification is the supervised-learning task of assigning a discrete label to an input, and different algorithms draw wildly different decision boundaries even on identical data. This simulator trains two of the most common ones — k-Nearest Neighbours and Logistic Regression — on the same 2D point cloud and renders each classifier's output probability as a literal 3D terrain: high ground means "confidently class 1", low ground means "confidently class 0". A movable threshold plane slices the terrain to show the exact decision boundary, instantly recolouring misclassified points, while a genuine ROC curve — swept across every threshold from the classifiers' real scores — shows the trade-off between true and false positive rates and reports the computed AUC. Adjust k, class separation and the threshold to see how a local, distance-based vote differs from a single global linear boundary.