Skip to main content
B
baraspatch1-Visitor
1-Visitor
February 15, 2018
Solved

Anayltics Dataset upload - empty string and empty value

  • February 15, 2018
  • 2 replies
  • 2199 views

Hey All,

I am trying to upload data into an dataset the analytics builder. I have identified that some of my data rows don't have values. 

Is there a way to make data optional in the configuration json file, or there a blank assumption when uploading into analytics that every single row and column has data. 

If this is the case then i must completely clean the data before upload, I do not want the OpType to "INFORMATIONAL" either. 

I understand the variances that might happen when uploading missing data , but curious if there an optional parameter that can be used.

Thanks in advance for the help

Cheers

Paul B

Best answer by cmorfin

Hi Paul

 

The best practice is to indeed prepare the data prior uploading it to Analytics Server.

This is a very  important step for all data science / machine learning tool. The end result model created is only as good as the data it trained on. If you have missing or inaccurate data, the prediction model generated will be poor.

This may represent a sizable amount of work, but well worth it as only the person you knows the data will fully understand how missing data should be repalced.

 

ThingWorx Analytics will attempt, in some cases, to do some transformation for missing value but this is unlikely to be as good as if you prepare the data yourself. 

https://www.ptc.com/en/support/article?n=CS228295 list a few rules that ThingWorx Analytics Server has got to transform missing data.

 

Hope this helps

Kind regards

Christophe

 

2 replies

C
cmorfin19-TanzaniteAnswer
19-Tanzanite
February 15, 2018

Hi Paul

 

The best practice is to indeed prepare the data prior uploading it to Analytics Server.

This is a very  important step for all data science / machine learning tool. The end result model created is only as good as the data it trained on. If you have missing or inaccurate data, the prediction model generated will be poor.

This may represent a sizable amount of work, but well worth it as only the person you knows the data will fully understand how missing data should be repalced.

 

ThingWorx Analytics will attempt, in some cases, to do some transformation for missing value but this is unlikely to be as good as if you prepare the data yourself. 

https://www.ptc.com/en/support/article?n=CS228295 list a few rules that ThingWorx Analytics Server has got to transform missing data.

 

Hope this helps

Kind regards

Christophe

 

L
LeighEmeritus
Emeritus
April 6, 2018

Hi @baraspatch

Just wanted to follow up to see if Christophe answered your question.  If so, please indicate Accepted Solution for the benefit of the Community.  If not, please advise on current status.

Thanks!

Leigh

Terms & ConditionsAccessibility statement