Classification with Neural networks

I am working on big data classification. I have tested many classifiers on big datasets. At the end, I have noticed that two classifiers achieved the best results but with long running time (Bagging & Random Tree). My question is: I want to modify Random Tree to use neural networks, but I want to know before I wast a lot of time on programming for nothing, which type of neural networks should I use for improving the classifier results and running time?

• Deep learning
• backpropagation
• stochastic gradient descent
• or a combination between them

=======

Thanks in advance for everyone in StackOverflow they are the best