RankingFeatureSelection/LambdaMART.py at master · bellal89/RankingFeatureSelection · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
import os
import math
from FeatureVector import FeatureVector
from RankFilter import RankFilter
import matplotlib.pyplot as plt

data_folder = "RankLib/Data"

def parse(line):
	parts = line.split('#')[0].strip().split(' ')
	rel = int(parts[0])
	qid = int(parts[1].split(':')[1])
	return FeatureVector(rel, qid, parts[2:])

def clear_folder(folder):
	for file_name in os.listdir(folder):
		full_file_name = os.path.join(folder, file_name)
		try:
			os.remove(full_file_name)
		except:
			print("Cannot delete file " + full_file_name)

def gen_plot(features_file_name):

	if os.path.exists(features_file_name + '.Weights'):
		weights = [float(w.strip()) for w in open(features_file_name + '.Weights', 'r').readlines() if w.strip() != '']
	else:
		f_f = open(features_file_name, 'r')
		feature_lines = f_f.readlines()
		f_f.close()

		feature_vectors = []
		for line in feature_lines:
			feature_vectors.append(parse(line))
		rank = RankFilter(feature_vectors)

		weights = rank.filter(3)

		fw_f = open(features_file_name + '.Weights', 'w')
		fw_f.write("\n".join([str(w) for w in weights]))
		fw_f.close()

	# print(weights)
	sorted_weights = sorted([(i + 1, w) for i, w in enumerate(weights)], key=lambda t: t[1], reverse=True)

	clear_folder("RankLib/SelectedFeatureSets")
	for i in xrange(1, len(sorted_weights) + 1):
		f = open("RankLib/SelectedFeatureSets/"+str(i), 'w')
		for w in sorted_weights[0:i]:
			f.write(str(w[0]) + '\n')
		f.close()

	clear_folder("RankLib/Results")
	x = []
	y = []
	for i in xrange(1, len(sorted_weights) + 1):
		score = get_test_ndcg(i)
		print(i)
		print(score)
		x.append(i)
		y.append(score)

	# Create the plot
	plt.plot(x, y, '-')
	plt.title('Relief-like algorithm')
	plt.xlabel('# of best features')
	plt.ylabel('NDCG@10')
	# plt.legend(features_file_name)
	# Save the figure in a separate file
	plt.savefig("RankLib/" + features_file_name + ".png")

def get_test_ndcg(features_count):
	cmd = "java -jar RankLib/RankLib.jar -train " + features_file_name + " -feature RankLib/SelectedFeatureSets/" + str(features_count) + " -ranker 6 -metric2t NDCG@10 -tts 0.65 -sparse > RankLib/Results/" + str(features_count)
	os.system(cmd)
	f = open("RankLib/Results/" + str(features_count), 'r')
	ndcg = float(f.readlines()[-1].split(':')[1].strip())
	f.close()
	return ndcg

for file_name in os.listdir(data_folder):
	full_file_name = os.path.join(data_folder, file_name)
	gen_plot(full_file_name)