Isolation forests appear to be one of the best options available for anomaly detection. I was disappointed to see that ML.NET doesn't support this algorithm.

I don't have much experience in this field, but these people do and they seem to agree: http://www.eurecom.fr/en/publication/5334/download/data-publi-5334_2.pdf

One important aspect of isolation forests is that they allow for unsupervised online learning. There are endless applications for such an algorithm. I'm hoping to see this available in ML.NET in the future!