Classification learner or neural network for supervised learning

Hi everyone, Beginners question here. I have a problem which I think I can identify as a classification problem which I hope to solve using supervised learning. I've had a good look through Mathworks documentation and elsewhere online in an attempt to answer the following question, without much luck.

Which tool should I use for this type of problem, the classification learner or a neural network from the Neural network toolbox?

My reading of this is that there seems to be some overlap in functionality. If the answer is 'it depends!!' what factors should I look at before choosing...if you have a reference online or in print, great! Thanks very much in advance for your assistance...and patience. regards Kevin

[Greg Heath]

Insufficient information. Please explain your problem.

[KB]

Thanks Greg, Perhaps I need to do a bit more digging before I start posting queries here. Essentially, I have data in the form of exam results in a range of subjects from a group of people who have applied to do a short course. I am trying to predict whether they will pass this short course based on their test scores on admission. I suppose the point is that I am new to machine learning and am attempting to halve my work by ruling out either Neural networks or the Classification learner. Maybe I need to try both and see which works best. OR maybe I need to withdraw the question and get a good book on machine learning and stop looking for shortcuts...

[Greg Heath]

In order to use a supervised classification NN, you need to train the net on previously classified examples.

If you don't have that, you could cluster (unsupervised classification) the applicants based on scores from different examples within the test.

However, I'm not exactly sure how you would use that info for 2-class classification.

Answer 1

[Sean de Wolski]

Maybe I need to try both and see which works best.

That's what I'd start with.

The next question becomes what are the requirements for your classifier? Is it accuracy? Prediction speed? Ease of interpretation? The things that are answered in the "Choose a classifier" documentation page (which doesn't discuss NN :( ).

[KB]

Thanks Sean and Greg, At least I have a starting point now....I'll give both routes a go and see how I get on. I suppose that is the beauty of MATLAB, you can explore these sort of things much more quickly than with other tools. Have a good day and thanks again for taking the time to answer. Kevin