I am working on some algorithms for anomaly detection

The dataset is clean our anomalies so I want to add some artificial anomalies.

I have added some anomalies. I get the maximum value of the dataset and add 20-25%, meaning these added anomalies are bigger than the max value by 20 to 25%

Are there any other types of anomalies that would be nice to have had in an anomaly detection algorithm dataset?

My dataset is with integers and float

  • $\begingroup$ What abount the minimum value? $\endgroup$ Apr 28 '20 at 20:53
  • $\begingroup$ Can you elaborate more on the structure of the data you have? Is it just a single list of integers/floats, or is each data instance an array? If each instance has multiple components (i.e. if each instance is more than a single number), there are many different types of anomalies you can produce just by thinking more about the one you already came up with. For each coordinate you can produce an anomaly which is "typical" in all coodinates except that particular one, and is extra large/small in that coordinate, using your max/min idea. $\endgroup$ Apr 28 '20 at 21:40
  • $\begingroup$ I have two columns, one column has the strings( description) the other column has the integers and the float data. These two columns are connected with each other meaning depending on what the string says the numbers on the other column can be different. What other anomalies would you suggest to add? $\endgroup$
    – E199504
    Apr 29 '20 at 7:35

There are many options. Here a couple:

  • Values that very high or very low.
  • Values that are many standard deviations from current distributions.
  • Multivariate anomalies - the combination of values across features are anomalous

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.