I have a few APIs that are called by clients. I collect data on them such as, what APIs they call, how often they call them etc. So, I have about 6 important metrics. I want to build an anomaly detection system around this. I have the following questions,
1) I want to build a simple model that can tell me if a new observation is an outlier based on the data I already have. Specifically, I am looking at implementing the low pass filter described in datascience.com/blog/python-anomaly-detection post under the simple statistical methods section. Are there any drawbacks of implementing it this way?
The reason I ask about drawbacks is because I mostly see SVMs being used to solve such problems. I don't want to get fancy with it if I don't have to. Again, I don't care about predicting the new value, I just care about identifying if it doesn't correspond with normal behavior.
2) Can the 3-sigma limits method be used when there are multiple features? I have only seen it be used for univariate data. For example, https://stackoverflow.com/questions/2303510/recommended-anomaly-detection-technique-for-simple-one-dimensional-scenario
If you all can also point to resources implementing similar problems, I'd appreciate it.