In this project I have played around with the famous KDD Cup 99 dataset and the challenge is to classify the network connections as ‘good’ or ‘bad’. Details of the project can be seen in the repo.
Generally people model algorithms on the 10% percent dataset, due to huge size, and apply it to full dataset. Thanks to Google Colab for the free GPUs I was able to run the nerual net from tensorflow 2.0 library on full dataset. Accuracy obtianed in 99.75