def trainModel(training_set, out, features=None, kcv=None, ranker=6,
leafs=10, trees=50, frate=1.0, shrinkage=0.1,
srate=1.0, bag=1, metric2t='DCG@10'):
"""
ranker
- 6 for LambdaMART
- 8 for RandomForest
RandomForest params
frate - what proportion of features are candidates at each split
srate - what proportion of the queries should be examined for each ensemble
"""
ranky_loc = check_for_rankymcrankface()
training_set_path = write_training_set(training_set)
cmd = 'java -jar {} -ranker {} -shrinkage {} -metric2t {} -tree {} -bag {} -leaf {} -frate {} -srate {} -train {} -save {} '.format(
ranky_loc, ranker, shrinkage, metric2t, trees, bag, leafs, frate, srate, training_set_path, out)
if features is not None:
import tempfile
features_file = os.path.join(tempfile.gettempdir(),'features.txt')
with open(features_file, 'w') as f:
f.write("\n".join([str(feature) for feature in features]))
cmd += " -feature {}".format(features_file)
if kcv is not None and kcv > 0:
cmd += " -kcv {} ".format(kcv)
print("Running %s" % cmd)
result = os.popen(cmd).read()
print("DONE")
return parse_training_log(result)
def trainModel(training,
out,
features=None,
kcv=None,
ranker=6,
leafs=10,
trees=50,
frate=1.0,
shrinkage=0.1,
srate=1.0,
bag=1,
metric2t='DCG@10'):
"""
ranker
- 6 for LambdaMART
- 8 for RandomForest
RandomForest params
frate - what proportion of features are candidates at each split
srate - what proportion of the queries should be examined for each ensemble
"""
cmd = 'java -jar data/RankyMcRankFace.jar -ranker {} -shrinkage {} -metric2t {} -tree {} -bag {} -leaf {} -frate {} -srate {} -train {} -save {} '.format(
ranker, shrinkage, metric2t, trees, bag, leafs, frate, srate, training,
out)
if features is not None:
with open('features.txt', 'w') as f:
f.write("\n".join([str(feature) for feature in features]))
cmd += " -feature features.txt "
if kcv is not None and kcv > 0:
cmd += " -kcv {} ".format(kcv)
print("Running %s" % cmd)
result = os.popen(cmd).read()
print("DONE")
return parse_training_log(result)
def trainModel(training,
out,
features=None,
kcv=None,
leafs=10,
trees=10,
metric2t='DCG@10'):
cmd = 'java -jar data/RankyMcRankFace.jar -ranker 6 -metric2t {} -tree {} -leaf {} -train {} -save {} '.format(
metric2t, leafs, trees, training, out)
if features is not None:
with open('features.txt', 'w') as f:
f.write("\n".join([str(feature) for feature in features]))
cmd += " -feature features.txt "
if kcv is not None and kcv > 0:
cmd += " -kcv {} ".format(kcv)
print("Running %s" % cmd)
result = os.popen(cmd).read()
print("DONE")
return parse_training_log(result)