如何使用Libsvm计算“准确性”-SVM_Predict.exe

Question

I am using the LIBSVM for the first time. I was able to train a data(for images) and my model is ready "trainingdata.svm.model"

Now, when I run my classification against an unknown test data it is giving me two files: 1. trainingdata.svm.prediction (This file contains 1's and 0's) against my each of test data. 2. It is giving me Accuracy = 8 %

THE QUESTION: 1. How do I interpret the 1s and 0s in my "trainingdata.svm.prediction". Note: I am classifying genders where 1 could be male and 0 could be female.

2. How is Accuracy calculated? How can a program calculate accuracy since the test data is an unknown entity and we do not know the labels yet.

Thanks

Answer 1

1. The file "trainingdata.svm.prediction" is predicting the labels 1 and 0 for your set (1 means the sample was predicted to be male, 0 is female).

2. It assumes all the labels belong to class index 0, I believe.