Drag to orbit ยท click a point to inspect its neighbours
โ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.
This lab renders a synthetic dataset as a 3D point cloud โ several dense clusters plus a handful of injected outliers โ and runs the real Local Outlier Factor algorithm over it: for every point it finds the k nearest neighbours, computes a local reachability density from their reach-distances, and scores each point by how much sparser its neighbourhood is than its neighbours' neighbourhoods. Points are coloured by that score from blue (LOF โ 1, a normal point sitting in a neighbourhood as dense as its neighbours') through yellow to red (LOF far above 1, a genuine local outlier), the k slider changes how many neighbours define "local", the outlier-injection and cluster-spread sliders let you stress-test the algorithm against harder datasets, and clicking any point draws lines to its k-nearest neighbours while reporting its exact LOF score.