The Machine Learning “Advent Calendar” Day 9: LOF in Excel

The Local Outlier Factor (LOF) algorithm for anomaly detection is explained through a step-by-step process using simple datasets in Excel. The method involves calculating k-distance to find a point's local radius, determining reachability distance to stabilize density comparisons, and finally computing the LOF score. By comparing LOF to Isolation Forest, it demonstrates that different unsupervised algorithms can yield different results on the same data. This highlights that the concept of an "anomaly" is not absolute but is defined by the specific logic of the model being used. Understanding the algorithm's mechanics is therefore more important than the final score it produces.