What is the ratio between input and output in RBF during testing?

I am doing my project in OCR.During training phase (using Radial Basis Function),i have got 96.85% accuracy.But during testing i have got poor results.I have used image that are written by same person for testing and training phase. Testing phase:[ feature vector size:879*48; target vector size:12*560 ] I want to know the ratio between feature vector(input) and target vector) during testing.

Training phase:[ feature vector size:879*560; target vector size:12*560 ]

I don't know how to improve the efficiency during testing . Could anybody help me to solve this issue...