Splunk for statistical testing/ Analysis

[jesslam]

Hi ,I am one beginner of splunk. May I know whether Splunk supports any statistical testing/ analysis, e.g. Normality tests, SVM, decision trees, Naïve Bayes, t-test, ANOVA etc. Thanks so much!

[imakaremi_splun]

Jesslam,

Since you asked this question, a lot has changed and new features have been added to Splunk.

Splunk's MLTS offers more than 20 advanced machine learning algorithms including SVM, Decision Trees, Naive Bayes, and many more. With its new SPL commands and custom visualizations, you can perform advanced machine learning analyses within Splunk.

Splunk's ML team is actively working on this app, adding new features, and making a lot of improvements to it.

[martin_mueller]

This might be right up your street: http://apps.splunk.com/app/1735/

[lguinn2 · Accepted solution]

No, Splunk does not calculate most of the statistics directly. Here are the basic statistics that you can calculate with Splunk:

Common stats functions

And there are commands other than stats that you can use, like contingency, that will either provide statistics directly or do part of the calculations.

Next, you can use the eval command to calculate almost anything. In addition to basic arithemetic, eval supports a number of functions:

Functions for Eval and Where

Finally, you can export the data from Splunk as a .csv file. So it is possible to calculate the basic statistics across millions of events, and then export those intermediate results so that you can continue the analysis with some other tool. You can even write code, such as Java or Python, to run Splunk searches and then analyze the resulting data however you like.