I'm using the Machine Learning Toolkit (MLTK) to detect outliers. It envelopes my line chart between the upper and lower bounds and uses these to determine whether or not there are outliers.
If I reduce the number of data points by zooming in on a particular time period of my line chart, the number of outliers increases. I know that the number of data points changes the numbers in the math, and that we have to tune our model to our needs, but do you have any advice on how to tune this, or how to determine what a large enough sample is so that I don't miss any outliers?
The major outliers are always obvious, but how can I make my outlier detection more "capable" at detecting outliers that are not so obvious (since they're not as drastically deviated as some other outliers are)?