The Machine Learning Advent Calendar Day 9: LOF in Excel

The article discusses the Local Outlier Factor (LOF) algorithm, illustrating its process through three steps: calculating distances and neighbors, determining reachability distances, and computing the final LOF score. By applying LOF to small datasets, it demonstrates how different algorithms may identify anomalies differently, emphasizing that in unsupervised learning, outlier definitions are subjective rather than absolute. This highlights the importance of understanding the underlying criteria used by various anomaly detection methods, as there is no single "true" outlier, but rather multiple valid perspectives based on the chosen algorithm and parameters.