After applying early stopping my test_accuracy decreases.
I have used 3 models
1)model (val_accuracy=82)
2)model1 with checkpoint (val_accuracy=85)
3)model2 with checkpoint+earlystopping(val_accuracy=81)
this is link https://drive.google.com/file/d/1s-69MkYk8x_LsdUCbrHnDvg_Fgi9ciEN/view?usp=sharing
Keras Callbacks - EarlyStopping,
This is because you have used value for attribute patience as 1, which means your model will stop training as soon as there is a small fall in val_acc. And the case is same here, your models val_acc fall a little from 81.20 to 80.84. So increase the patience value to any integer from 3-5 and than see the results.
Hope this cleared your doubt.
ok i will do .but i think sir also used patience as 1 ,but accuracy increased
I got accuracy as 0.838919997215271 for patience=4
Everytime keras initializes different weights to the layers which are not the same, that is the reason model training differ on different instances, this is the answer of statement in first comment.
Fair enough, these accuracies are close enough. So the results are acceptable.
Happy Learning 
I hope I’ve cleared your doubt. I ask you to please rate your experience here
Your feedback is very important. It helps us improve our platform and hence provide you
the learning experience you deserve.
On the off chance, you still have some questions or not find the answers satisfactory, you may reopen
the doubt.