How is the output from a trained model to be understood?

1 次查看（过去 30 天）

Betzalel Fialkoff 2018-3-25

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/390580-how-is-the-output-from-a-trained-model-to-be-understood

编辑： Betzalel Fialkoff 2018-3-25

So I trained a logistic regression model, using the Classification Learner app, after training I exported using the generate code option, at this point I have the model in my workspace as a struct. My code is as follows:

%feature extraction
features=ReadData();
%train the model
logitReg=trainingModel(features);
%to predict create a feature vector and call the predictFcn method of the
%model
featureVector=[0 0.01 0.1 1 2];%A random feature vector to test the function
[a b]=logitReg.predictFcn( featureVector);

My problem is simply that I dont know how to understand the output. As far as I can tell the variable a is the model's prediction, b is a 2x1 double, the sum of its 2 terms is equal to one. In this case:

b=[0.0239 0.9761];

this leads me to believe that b(1) is the probability that the feature belongs to class '0' and b(2) is the probability that the feature belongs to class '1'. However I cannot find ny documentation to confirm this.

Can anyone help? Greatly appreciated!