03 March 2018 0 4K Report

I am working on a deep neural net model to estimate the monthly mortgage prepayment ratio for a large loan portfolio (this ratio is called SMMSMM in finance literature, all you need to know is that 0≤SMM≤10≤SMM≤1). My basic approach is to extend the methodology in (Papke- Wooldridge 1996) paper on modelling fractional outcomes. So I started by

  • replacing the G(z)G(z) (sigmoid function in the above paper) by a deep NN.
  • minimizing the negative Bernoulli log-likelihood function: argminθ(−∑iyilog[G(Xiθ)]+(1−yi)log[1−G(Xiθ)])argminθ(−∑iyilog⁡[G(Xiθ)]+(1−yi)log⁡[1−G(Xiθ)])
  • My question is: what is the conceptual difference between using the above cost function as opposed to the usual Mean Squared Error cost for a regular quantitative response 0≤y≤10≤y≤1?

    Similar questions and discussions