How to increase classifier accuracy on evaluation data

Question

I have trained a classifer that has decent training and validation accuracy and loss.

Training accuracy: 0.928
Training loss: 1.949
Validation accuracy: 0.920