Sklearn SVM makes incorrect boundary line on a clearly linearly seperable dataset

Question

I\'m currently working on my thesis about fault detection using SVM. So I used kernel fisher discriminant analysis to extract 52 features of my dataset into plot-able 2 features