Splunk Search

Why does predict command not predict?

perryd
Engager

Hi everyone, i need help to understand why i'm wrong and how to fix the problem.

I've a lookup table in which is stored the last four year of data. That have a seasonality of a month and i want to predict the next year.

 

I use the predict command with the LLP algorithm to estimante the values. Below i show you my query and the output:

perryd_0-1655474384774.png

As you can see, the prediction doesn't work because it simply put the last two values in ciclic way.

here the associated table:

perryd_1-1655474480897.png

could you help me to understand where i'm wrong with query or data?

 

i did the same work months ago with different output, more realistic:

perryd_2-1655474786408.png

Thaks a lot!

Labels (2)
Tags (1)
0 Karma

bethj
New Member

Did you find out why predict was not working for you? 

I have been running into the same issue. Everything was predicting correctly until earlier this year. Now, it is just providing identical forecasting, even though nothing has changed on our side.

0 Karma

isoutamo
SplunkTrust
SplunkTrust

Hi

why you are thinking that it's wrongly? Based on your chart it seems that your current data have different values from where predict do the prediction. As you can see the latest month has much lower value than you have in previous month. I suppose that this will make the difference between your monthly predictions.

Basically LLP should fix this, but as (based on your charts) you haven't enough data points (The number of data points must be at least twice the number of periods, using the period attribute.) this probably didn't work correctly with the seasonality what you have with the last month (trx sum has decreased really much)?

r. Ismo

0 Karma

perryd
Engager

Hi Ismo,

in my post, dataset was an example, here my dataset: 

perryd_1-1655823114648.png

 

i have 53 months, so over 4 years splitted per months and in relation of documentation, the period is assumed by span parameter in timechart.

here i've the same "problem". I'm using LLP algorithm because i've a seasonality. As you can see, in August more or less, i've a drecreasing of volumes and then an increasing. The prediction instead show a "zig zag" trend with the two same values. 

In this way, the prediction is unuseful because i'm expecting a decreasing in this period and i'm not able to understand the future trends. Is my data the problem? too hard to predict and to establish a trend?

Thanks,

Denis

 

 

0 Karma

isoutamo
SplunkTrust
SplunkTrust

Do you remember that LLP is not for trend, it’s just for predicting with seasonality. Based on that chart I suppose that your data is not suitable for getting the best estimate? Have you cleaning up it or is it just timeseries data from raw with some calculations?

Here are couple of articles which may explain that issue?

0 Karma
Get Updates on the Splunk Community!

Index This | I am a number, but when you add ‘G’ to me, I go away. What number am I?

March 2024 Edition Hayyy Splunk Education Enthusiasts and the Eternally Curious!  We’re back with another ...

What’s New in Splunk App for PCI Compliance 5.3.1?

The Splunk App for PCI Compliance allows customers to extend the power of their existing Splunk solution with ...

Extending Observability Content to Splunk Cloud

Register to join us !   In this Extending Observability Content to Splunk Cloud Tech Talk, you'll see how to ...