A Keras model to predict who will win Donald Trump or Hillary Clinton in a county depending on various features like 'Per capita income' , 'Total hoseholds' , 'Unemployemnt rate' , etc. in that county.
Datasets used are in the folder 'data'.
1.) Keras with Tensorflow backend
2.) Numpy
3.) Pandas
To execute the code, just run the main.py file. It will call the required model for training and testing.
Statistics at Epoch 100/100: loss: 0.2149, acc: 0.7580, val_loss: 0.2139, val_acc: 0.8361
It is observed that increasing the number of epochs increases accuracy and further decreases loss.
The dropouts also allowed significant improvement (by several %). This is because with using Dropouts, I was able to avoid overfitting.
I also set differing number of units when calling Dense() on each layer. This adjusted the dimension of the output. I noticed that when I steadily increase, then decrease the number of units across my layers the accuracy improved.
Through these changes and adjustments, my validation accuracy went up to around 83% with 100 epochs. It is important to note that the training accuracy stayed at around 76%. This is due to the Droputs. Therefore, it is a good sign that the validation accuracy is higher than the training accuracy: this proves that Dropouts alone improve the accuracy by several %.
