train = train[random_select, :]
labels = labels[random_select]
# set up the model
params = {'n_estimators': n_estimators,
'learning_rate': learning_rate,
'algorithm': algorithm}
clf = ensemble.AdaBoostClassifier(n_estimators=params['n_estimators'],
learning_rate=params['learning_rate'],
algorithm=params['algorithm'])
# write parameters
logfile = 'log_adb/adb_cv_%d_%f_%s' %(params['n_estimators'],
params['learning_rate'],
params['algorithm'])
pname = logfile+'.txt'
xu.write_params(pname, clf.get_params())
# write results
fscores = open(logfile+'.csv', 'w')
# run CV
scores = cross_validation.cross_val_score(clf, train, labels,
cv=2, scoring=logloss_mc,
verbose=0)
print "scores: ", scores
fscores.write('%f %f\n' %(scores.mean(), scores.std()))
fscores.close()
'bootstrap': True, 'n_jobs': -1, 'verbose': 0}
clf = ensemble.RandomForestClassifier(n_jobs=params['n_jobs'],
n_estimators=params['n_estimators'],
max_depth=params['max_depth'],
max_features=params['max_features'],
verbose=params['verbose'],
warm_start=True)
# oob_score=True)
print clf.get_params()
fscores = open('log_rf/scores_%d.csv' % arg1, 'a+', 1) # line buffering
for i in range(1, params['n_estimators'] + 1):
clf.set_params(n_estimators=i)
scores = cross_validation.cross_val_score(clf, train, labels,
cv=2, scoring=logloss_mc,
verbose=0)
print "CV #: %d" % i, "scores: ", scores
fscores.write('%d %f %f\n' % (i, scores.mean(), scores.std()))
fscores.close()
# scores = cross_validation.cross_val_score(clf, train, labels, cv=5,
# scoring=logloss_mc, verbose=1)
# np.savetxt('log_rf/scores.csv', [scores.mean(), scores.std()])
fname = 'log_rf/params.txt'
xu.write_params(fname + '.txt', clf.get_params())
