# ShelterModel Solution for Kaggel [Shelter Animal Outcomes](https://www.kaggle.com/c/shelter-animal-outcomes) This is a Repo source the Clean Function from [Mike Fang](https://github.com/fhlgood/K_sa/blob/master/clean_original.R) and usesfeaturing introducing Sizes of dogs and breaking Dogs Breeds combinations. The logloss was 0.82. Worse than the best LB that was 0.72 but taking account time and hours is nonsense Because it have to much predicitve value but too low practical value. So i will not work any more. This solutions includes a list of Sizes by breed. The original worksheet for sizes can be find editable in this [link](https://docs.google.com/spreadsheets/d/1yTCvdXgY0JLYNfYd4GdLphxMgnur81TBbtD2nwRdcHc/edit#gid=0) The competition had have more intrinsic data about animals. Weight lenght, health conditions. etc. Althought our size information does not improve variability too much. Mixing with information of dogs groups(Terrier, toy, sport) may have improvement in the model.  