Neural network as challenger in classification tasks

Original by DataQuest, 2017, 6 pagesHamster_gagarin_linkedin
hamster writter This summary note was posted on 21 July 2018, by in Credit risk #, #
  • Can use single neuron to do logistic regression (use logit as a smooth function)
  • Use python’s Kera package (
  • Challenge model with a small neural network of two connected hidden layers with each 10 neurons
  • Identify possible non linearities by looking at default rate occurrence per driver
  • Examine gini as a function of driver threshold
  • For bigger datasets use Python Theano or Tensorflow