sdevkota007/Glass-Classification-With-SVM

Classification with SVM for Glass Data Set from UCI

Glass-Classification-With-SVM

This repository contains implementation of SVM for classification of glass. The dataset can be found here https://archive.ics.uci.edu/ml/datasets/glass+identification
To achieve the best classification accuracy, four different types of kernels were tested: RBF, linear, polynomial and sigmoid. In addition to accuracy, the training time was also compared for two different classifier type: OneVsOne and OneVsAll classifiers
Here,
ovo-balanced means OneVsOne classifier with balanced class weight
ovr-balanced means OneVsRest classifier with balanced class weight
ovo-balanced means OneVsOne classifier with unbalanced class weight
ovR-balanced means OneVsRest classifier with unbalanced class weight

Results:

Training Time:

Accuracy table for 5 fold cross validation:

ovr-unbalanced

ovr-balanced

ovo-unbalanced

ovo-balanced