All Apps and Add-ons

Machine Learning Toolkit: Why is Predication showcase not working as expected?

Engager

Hi All,

This is regarding Predication showcase of Machine Learning Toolkit App. I am using License usage query, there are three fields "Info", "Warning" and "Critical". Please find the current fields value.

Info : More than 700 GB and less than 800 GB
Warning : 0
Critical: 0

Prediction Query : | fit LinearRegression fit_intercept=true "Info" from "_time" "Critical" "Warning" into "default_model_name".

"Info" field is used for prediction. "Warning", "Critical" and "_time" fields are used for predicting the value. I am not getting the corresponding actual prediction result. Please check and let us know to proceed further.

0 Karma

Splunk Employee
Splunk Employee

Is the content of the Info field literally the text you provide in the example? If so, then please note that LinearRegression is used to predict the value of a numeric field and the sentence "More than 700 GB and less than 800 GB" is not a number. (Even if the field contains "700GB", that will still be treated as a string.) Could you confirm that Info contains only "700" or "800" or something similar?

0 Karma

Splunk Employee
Splunk Employee

"700 to 800" is not numeric; "700" and "800" are numeric. Let's assume you have the latter. Could you please give a concrete example of an input, the result produced by fit, and the result you expected? A screenshot might be helpful, too.

0 Karma

Engager

It ranges from "700" to "800".
Please find the Inputs.
_time Info Warning Critical
2016-12-04 0.00000 0 0
2016-12-05 724.14825 0 0
2016-12-06 772.61592 0 0
2016-12-07 789.88528 0 0
2016-12-08 774.01143 0 0
2016-12-09 751.44294 0 0
2016-12-10 581.31427 0 0
2016-12-11 560.07010 0 0
2016-12-12 695.39823 0 0
2016-12-13 733.44824 0 0

Please find the Prediction results.
Info predicted(Info) residual _time Critical Warning
724.14825 605.99 118.16 05-12-2016 0 0
772.61592 607.35 165.27 06-12-2016 0 0
560.0701 614.18 -54.11 11-12-2016 0 0
733.44824 616.91 116.54 13-12-2016 0 0
696.90872 621.01 75.9 16-12-2016 0 0
731.71623 626.47 105.25 20-12-2016 0 0
690.88541 629.2 61.69 22-12-2016 0 0
656.51752 630.56 25.96 23-12-2016 0 0
601.12301 634.66 -33.54 26-12-2016 0 0
676.66416 637.39 39.27 28-12-2016 0 0

The prediction result is continuously increasing from 600. There should be up and down in the results as per the input. we are expecting a approx. results based on the inputs.

0 Karma

Engager

All fields are numeric including Info filed. "GB" was just to specify them as we have used the License usage query.

Its "700 to 800".

0 Karma