I tried to implement outlier detection for one dimension using inter quartile. For instance, a given variable cost or revenue or profit. but I'm missing outliers in other dimensions when running for one dimension.

How to detect outliers for multi dimensional data , like cost and revenue and profit at once.

Is there any efficient algorithm for this?

  • 1
    $\begingroup$ It depends on your purposes. You can take a look at Mahalanobis distance if your data is normal. $\endgroup$ – German Demidov Feb 23 '16 at 12:04
  • $\begingroup$ Unfortunately data isn't normally distributed $\endgroup$ – tourist Feb 23 '16 at 12:14
  • $\begingroup$ Then you can use combined p-values: en.wikipedia.org/wiki/Fisher%27s_method , but it will work only if you have independent coordinates. $\endgroup$ – German Demidov Feb 23 '16 at 12:23

If you are working on Python, you could try SVM One Class and Least Squares Anomaly Detection, both are unsupervised learning, so you train giving "examples" of inliers.


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