Calculate the evaluation metric "AMS" of Higgs Boson competition at Kaggle https://www.kaggle.com/c/higgs-boson/details/evaluation

calculate_ams.py

# Calculate the evaluation metric "AMS" of Higgs Boson competition at Kaggle 
# https://www.kaggle.com/c/higgs-boson/details/evaluation

"""
__Author__ = "Aziz Alto"
"""

import numpy as np
import pandas as pd

# input solution and predicted result files
true_sol = './solution.csv'			# assumes columns: EventId, Class, Weight
pred_res = './result.csv'			# assumes columns: EventId, RankOrder, Class

# read files
sol = pd.read_csv(true_sol)
res = pd.read_csv(pred_res)

# read Weight 'w' and Class(label) 'y' of the true solution
w = sol.Weight.values
y = sol.Class.values

# read predicted Class(label) 'p' of the result
p = res.Class.values

# sum up weights of s (True Positive) and b (Flase Positive)
s, b = 0, 0 					# initialize s and b
for w, y, p in zip(w, y, p):
	if (y == p == 's'):
		s += w 				#  s = sum(w_i * (y_i == 's') * (y^_i == 's'))
	elif (y == 'b') and (p == 's'):
		b += w 				#  b = sum(w_i * (y_i == 'b') * (y^_i == 's'))

# calculate AMS
def ams(s, b, b_r=10):
	""" b_r: constant regularization term """
	a = np.sqrt( 2 * ( (s + b + b_r) * np.log ( 1 + s / (b + b_r) ) - s ) )
	return a

print("s: {}, b: {}".format(s, b))
print("AMS: {}".format(ams(s, b)))