Hi Opeyemi,

If you need to ensemble several learning methods, Stacking and Vote algorithms are useful for this purpose. Hence, they should perform learning simultaneously under evaluation method such as cross-validation.

HTH.

Samer

ANN works on error-back-propagation-feed-forward technique where one can go on improving training the network. Whereas SVM works on structural risk minimization principle. These two may be possible to hybridize. However if you want to further improve training the network, one can do , ie, from ANN to ANFIS, ANFIS-PSO, GA-ANN; and from SVM to SVM-PSO, GA-SVM, etc. from ANN/SVM hydridizing with GA/ACO/PSO (which are optimization tools), those hybrid models should perform more efficient than ANN/SVM. You can see many applications of these models in ELSEVIER Journals - Applied Soft-Computing, Advances in Engineering Software, etc.

You can using the Stacking method, please have a look at this:

http://machine-learning.martinsewell.com/ensembles/stacking/

Hi Opeyemi,

Based on my little experience you can use more than one classifier/method in an ensemble method.

You can have a look at this:

http://link.springer.com/article/10.1007/s10462-009-9124-7

Regards

You might be interested in some ideas coming out from this paper.

Article Hybrid regression model for near-real time urban water deman...

You could a generative hyper-heuristics. The hybridisation can be achieved by generating automatically new metaheuristics from a learning set of instances. 

No, it will not give good results and computational complexity will be increased un-necessarily.