This is based on a simple competition conducted by Credit Suisse on Interviewbit
Dataset is on the various things a soldier did on the battlefield and, the respect is to be calculated between the scale of 0-1.
I've made use of simple Neural Networks, with six layers in combination of different initializers like he_uniform, lecun_uniform, etc.. a simple RELU as activation function with a constant training and playing with hyper parameters, I was able to achieve a 94.6% of accuracy on the test dataset
##the dataset is quite big, with over 40Lac reviews for training and over 1Lac for testing.