Hello, I need your suggestion...

I am doing some biomedical project. It is a calssification problem.

There are two groups. One is 'Normal' and other is 'Cancer'.

I have taken 30 subjects which is the mixture of 'Normal' and 'Cancer' for the analysis.

I used a concept called "Recurrence Quantification" to get features. In that I got DWT coefficient. I applied Recurrence analysis for DWT coffiecient and computed 6 values RR,DET,LMAX,ENT,LAM,TT.

My problem shown in figure. alt text

This is the feature set I got by Rrecurrence Analysis.

6 features are RR,DET,LMAX,ENT,LAM,TT which I got for DWT coefficients (An, D1 - D7, Global) (9 rows)

Coming to my classification problem,

So I got 9x6 matrix for each subject in 30 peoples group. So I need an algorithm such that It has to classify these two groups from the mixture of 'Cancer' and 'Normal' by using these 9x6 matrix. Which algorithm is better?

I heard that feature means values should be in vector from only. But mine is a 9x6 matrix. Then how can solve this?

Can we do this by LDA or SVM or any other... suggest me...

Thanks you guys!

asked Jan 27 at 10:18

Rick%20Martin's gravatar image

Rick Martin
0345

edited Jan 27 at 10:19

So you only have 30 training examples total? Also, are they labeled with the desired output (Normal vs. Cancer)?

(Jan 27 at 16:17) Joseph Turian ♦♦

No, I have to classify according to the values I got (9x6), instead of lables.

(Jan 27 at 17:50) Rick Martin

You want a classifier to predict the output label from the input values (9x6). Typically, one has examples of correct output labels for the given input values.

If you don't have any examples of correct output labels, then how are you going to know if your classifier works or not?

(Jan 27 at 18:13) Joseph Turian ♦♦

Ok, what algorithm works better, if we have labels?

(Jan 27 at 18:25) Rick Martin

How many labels would you have? 30? 1000?

(Jan 27 at 18:36) Joseph Turian ♦♦

There are 30 people(subjects) in the whole experiment. In which 15 are 'Normal' and other 15 are 'Cancer'. So, Two labels...

For each subject we get 9x6 matrix with corresponding label 'Normal' or 'Cancer'.

(Jan 27 at 18:50) Rick Martin

@Joseph Turian, Am I wrong with this interpretation ? Help me...

(Jan 27 at 22:44) Rick Martin
showing 5 of 7 show all
Be the first one to answer this question!
toggle preview

powered by OSQA

User submitted content is under Creative Commons: Attribution - Share Alike; Other things copyright (C) 2010, MetaOptimize LLC.