Skip to main content
1-Visitor
January 25, 2019
Solved

what happens when an analytics job is successful and how can the analytics model be improved

  • January 25, 2019
  • 3 replies
  • 1852 views

I am using an analysis provider to classify incoming events into 3 different cases/classes. how can I improve my model if I have less data?

Best answer by jgreiner

Hello,

 

After an Analytics job has completed successfully, a new predictive model is created.  You can create new models and access ones that were already created under the Models tab in ThingWorx Analytics Builder.  Post-Pic-1.PNG

 

Double clicking on a finished model will open the models results page which will give you different details about the model such as the Prediction Accuracy, RMSE, etc. 

 

Post-Pic-2.PNG

One way you can try to improve your models is by creating several different models and editing the Learning Techniques, Ensemble Techniques, and any other settings differently for each model until you get a model with the desired accuracy.

 

Post-Pic-3.PNG

As for your question on less data, sometimes it can be very difficult to create an accurate model if there is not enough data present.  We always recommend collecting as much data as possible and to continue adding that data to your datasets.   Usually with machine learning, the more examples you have, the more accurate the model is.

 

Warm Regards,

 

John

3 replies

jgreiner13-AquamarineAnswer
13-Aquamarine
January 25, 2019

Hello,

 

After an Analytics job has completed successfully, a new predictive model is created.  You can create new models and access ones that were already created under the Models tab in ThingWorx Analytics Builder.  Post-Pic-1.PNG

 

Double clicking on a finished model will open the models results page which will give you different details about the model such as the Prediction Accuracy, RMSE, etc. 

 

Post-Pic-2.PNG

One way you can try to improve your models is by creating several different models and editing the Learning Techniques, Ensemble Techniques, and any other settings differently for each model until you get a model with the desired accuracy.

 

Post-Pic-3.PNG

As for your question on less data, sometimes it can be very difficult to create an accurate model if there is not enough data present.  We always recommend collecting as much data as possible and to continue adding that data to your datasets.   Usually with machine learning, the more examples you have, the more accurate the model is.

 

Warm Regards,

 

John

mahitejg1-VisitorAuthor
1-Visitor
January 29, 2019

hi @jgreiner  does this creation of new model happens automatically or we have to create a new one and is there a place i can download all the result set data from hosted server trial

13-Aquamarine
January 29, 2019

Hi,

 

A new model must be created either by the user or a program that the user has created.  You can download the results file by selecting MODELS>Selecting the desired model>Predictive Scoring>Selecting the Download link for the desired scoring job.

 

Post-Pic-4.PNG

 

Thanks,

 

John