Accuracy Information in Feature Manipulation in AML Toolkit

Updated Feb 21, 2018

Reported In

Software

  • LabVIEW Analytics and Machine Learning Toolkit

Issue Details

Using the Feature Manipulation (Training).vi in the Analytics and Machine Learning Toolkit, I can reduce the number of features and save the model to a .json file. I can open this trained model in Feature Manipulation (Deployment).vi for testing, however I cannot get reading of Accuracy and Precision as in the Classification (Deployment).vi example. Is there a way I can read the Accuracy and other results from reduced feature models?
 

Solution

Accuracy and precision information cannot be obtained in the feature manipulation stage of the training. However, feature manipulation should be followed by classification in the model training. Accuracy information can be obtained through the Evaluate Classification Model VI in the classification stage. 
 

Additional Information

The process of training a model includes loading data into it which is followed by feature reduction and the classification. Feature reduction should always be done before classification. See figure below for training stages. 

The accuracy and precision obtained in the Classification VIs is calculated by the Evaluate Classification Model VI. This VI uses functions which compare the the model labels to the actual labels and it is done through a confusion matrix that is by definition only used in the classification stage. 

 

WAS THIS ARTICLE HELPFUL?

Not Helpful