Feature Weights in Models in Analytics and Machine Learning Toolkit

Updated Feb 14, 2018

Reported In

Software

  • LabVIEW
  • LabVIEW Analytics and Machine Learning Toolkit

Issue Details

I am using the Classification VIs in the Analytics and Machine Learning (AML) toolkit for LabVIEW. The model is saved into model.json using a number of features. How do I find out how much weight is each feature contributing to the final model?

Solution

Whether weight information of each feature can be extracted depends on the classification model. 

If a linear model, i.e. linear regression model, is used you can find the features' weights by looking at the factor for each variable as shown in the block diagram below. 
If a nonlinear model is used, like SVM or NN, it becomes more complex to extract this information because a lot of transformation and calculation is done on the input feature. There won't be any obvious indicators of each feature's weight. 

WAS THIS ARTICLE HELPFUL?

Not Helpful