Seeking Advises on how to perform One Class classification on a text data

2 visualizaciones (últimos 30 días)
Hi
I have some doubts on implementation of one class classifier . This question is trying to explain how am I implementing the one class classifier and I hope you will advice me whether I am approaching right way or not.
I am experimenting on applying one class classification upon a text dataset with 2 classes (class 1 and class 0).It is a balanced dataset. Intially my train matrix is of size 3000x100. After eliminating a class for the purpose of training my model for one class, it's reduced to 1500x100 (contains only class 1). And the test/validation matrix is of size 2000x100 where data of 2 classes exists (contains data for class 1 and for class 0).
To train the model, I am applying the following codeblock upon train data and have tried to calculate the precision, recall and F1:
model=fitcsvm(ReducedDataset,ReducedClassSet,'Standardize',true,'OutlierFraction',0.05);
predTrain = predict(model,ReducedDataset);
[recall,precision,f1]= scores(ReducedClassSet,predTrain);
where, ReducedDataset and ReducedClassSet represents the train matrix and classes. The scores I have achieved are 1, 1 and 1. After applying the model on the test data
prediction= predict(model,ReducedTestDataset);
[recall,precision,F1]= scores(classTest, prediction);
I get the scores of 1,0.5004 and NaN( which seems illogical to me).
My question is am I following the proper way of applying one class classifier? If am not, what should be the right approach then?
I am looking for your advice, in this regard.
Thanks,

Respuestas (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by