You have an Anomaly Detection project you're working on and you need a simple approach of clustering data into classified groups. Which algorithm is the best choice given this situation?

Anomaly detection can start with an inexpensive, unsupervised clustering technique. K-Means partitions the data into k compact clusters; any point lying far from every centroid can be flagged as an outlier. The algorithm is easy to implement, fast to run, and widely recommended as a baseline for anomaly-detection use cases.

B. Neural networks are mainly supervised, require heavy tuning, and are not purpose-built for clustering anomalies.
C. Decision trees classify labelled data; they do not form clusters or directly expose outliers.
D. Hidden Markov Models model time-series states, not general-purpose clustering for anomaly screening.

1. Bishop, C. M. Pattern Recognition and Machine Learning, Springer, 2006, Ch. 9.1, pp. 423-425.

2. Chandola, V. et al. “Anomaly Detection: A Survey,” ACM CSUR 41(3), 2009, §4.2, doi:10.1145/1541880.1541882.

3. MIT OCW 6.036 Intro to ML, Lecture 12 “Clustering & k-means,” slide 5 (2022).