how to tell which feature is making trouble in classification for getting better accuracy?

I'm working with image processing.I've trained my data with classification learner app.but i could not get expected accuracy in classifier.so is there any way to understand which feature/column is creating trouble for better accuracy?

Answer 1

[Image Analyst]

You can try doing PCA on it. It's possible some measurements are essentially just garbage/noise and not contributing anything worthwhile to the model. Just look at the amount of variance explained by the different PCs and which original variables have higher weights in the junk PC's than in the good PC's.

[Rashidul Alim Anik]

Is there any tool/app for doing PCA?

[Walter Roberson]

https://www.mathworks.com/help/stats/pca.html

[Image Analyst]

In addition to the link Walter gave (show hidden comments), there is also a PCA option in Regression Learner.