is it possible to get 100% classification accuracy in SVM classifier lerner app in matlab?

I am getting 100% accuracy in almost every model in classifier leraner app MATLAB . Initially i gt few NaN entries in ma feature set ,but after removing them i m still getting it same almost 5000 samples and 300 features.

Plz suggest some debug methods.

[Walter Roberson]

100% is possible if it was able to find a hyperplane that completely separated the data.

Try generating some random data from one distribution and assign randomly to classes, and see if it still claims 100%

[kc]

Sir i have divident the data randomely in 70-30 . Given 70 with labels in Classifier app, and then extract the model to test the remaining 30 , which is not given to classifier before , and still its giving high accuracy(almost same a lil lower). But i know on that dataset its almost impossible with these features. So plz guide what else i need to check?

one more important question is is it fine to use this app or i should go for code. But eventually that is same as generated code from app ? right?

plz help

[Image Analyst]

Please post a scatterplot of your data so we can see if the data points are well separated or not. Attach your data if you need more help.

Answer 1

[kc]

Answer 2

[kc]

Sir i have attched the scatter plot but data file is large so unable to do so.