cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
Showing results for 
Search instead for 
Did you mean: 

Analytics tutorial - Inaccurate model

SOLVED
Highlighted
Regular Member

Analytics tutorial - Inaccurate model

Hello,

I'm doing the tutorial to build an Analytical Model. The steps are quick and easy and yet I've got an innacurate model at the end.

Here is the dataset provided in the tutorial. I create a model with the basic configuration using all indicators with low_grease boolean as goal (same as the tutorial).

I don't undestand what I'm doing wrong. Any clue?

 

Tutorial result :

27_ab_models_view_confusion_matrix

 

My result :

Sans titre.png

 

Philippe

1 ACCEPTED SOLUTION

Accepted Solutions
Highlighted

Re: Analytics tutorial - Innacurtae model

Hola Philippe,

I have encountered this issue before, and for me the clue was to update my TWA instance.

This is what I get with 8.5.4

2020-05-13_17-11-31.png

Regards

View solution in original post

6 REPLIES 6
Highlighted

Re: Analytics tutorial - Innacurtae model

@plpt 

Thank you for opening the post on community. It seems there is major difference between actual model and your model.

I can see lot of values falling under False positive for you as per confusion matrix. Did you exclude any field while creating model?

Have you checked the profiling?

 

Regards,

Mohit

 

Highlighted

Re: Analytics tutorial - Innacurtae model

I did not exclude any field in my model.

What do you mean by check the profiling?

But the thing is that I reproduce every step of the tutorial...

 

Best regards,

Philippe

Highlighted

Re: Analytics tutorial - Innacurtae model

@plpt 

 

Signals and profiles is the 6th step in that tutorial. I am not sure why you are getting this inaccurate model.

 

Could you please try to retrain it or build and new model with same data set.

 

Also do try with Refine model steps given in 7th point and let me know if it works for you.

Highlighted

Re: Analytics tutorial - Innacurtae model

Profiles and Signals are the same than the tutorial, no trouble here.

 

When I retrain or rebuild a new model with the dataset the results are still around 0.4 for the precision, still bad.

 

Refining the model doesn't help either.

 

Could you tell me if you have the same result as the tutorial please?

 

Best regards,

Philippe

Highlighted

Re: Analytics tutorial - Innacurtae model

Hola Philippe,

I have encountered this issue before, and for me the clue was to update my TWA instance.

This is what I get with 8.5.4

2020-05-13_17-11-31.png

Regards

View solution in original post

Highlighted

Re: Analytics tutorial - Innacurtae model

It works! Thanks a lot!

Announcements

Thingworx Navigate content has a new home! Click here to access the new Thingworx Navigate forum!