heritage_health_base_model.py

import numpy as np

def rmlse(outputs, targets):
  targets = np.array(targets)
  outputs = np.array(outputs)
  log_diff = (np.log(outputs+1.) - np.log(targets+1.))
  rmse_val = np.sqrt(np.sum((log_diff ** 2))/len(outputs))
  print("rmlse is: " + str(rmse_val))
  return rmse_val

# find data here https://github.com/sandro-pasquali/RBelgium_HeritageHealthPrize/tree/master/Data
y2_days = [int(i.split(',')[-1]) for i in open('./raw_data/DaysInHospital_Y2.csv').readlines()[1:]]
y3_days = [int(i.split(',')[-1]) for i in open('./raw_data/DaysInHospital_Y3.csv').readlines()[1:]]
base_model_0 = [0]*len(y3_days)
base_model_avglog = [np.average(np.log(np.array(y2_days)+1.))]*len(y3_days)

rmlse(base_model_0, y3_days)
rmlse(base_model_avglog, y3_days)