Accuracy is not that good

#1

Here is my code:

I used image augmentation first because there are lesser number of images for each Pokémon. After augmentation, I trained the model having 5 layers but still I am not able to get certain accuracy.
I need to know that whether it is because I am using ANN over image classification or there is still anything can be done in order to increase the accuracy.

#2

hey @Joy-Gupta-2763139277246091 ,
Firstly , A really good notebook. Well done.
Secondly coming to your doubt , see improving a deep learning model doesn’t means that we need to always increase the number of layers in it.
It that was to be true , then we would be taking hundreds of layers to get it done.
As you are making it from scratch , in that case choosing the correct number of layers and deciding number of nodes in each layer plays a very important role.
Along it also matters how do you preprocess the input images and provide to your model.
Optimizing learning rate , number of epochs plays a very big role.

While dealing with images , in deep learning we use convolutional layers but to implement them from scratch is a quite tricky task and takes a lot of time.
You can try other models also , like KNN , or LogisticRegression . But no one can’t gurantee that which mode will work the best.
You need to try as much as you can and select the one that works best.

I hope this helps you :slightly_smiling_face:.

1 Like
#3

Thanks for the complement. And lastly, thanks for the suggestion. I’ll try to find the most accurate hyperparameter. And yeah will implement KNN and Logistic Regression too.

#4

Yeah , try once those too.

#6

hey @Joy-Gupta-2763139277246091 ,
I guess you have mistakenly shared a different notebook here.
This doubt was related to pokemon classification. and you have shared the boston house prediction code.

Kindly look to it.

#7

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.