),),
#('scaler' ,StandardScaler(copy=True,with_mean=False,with_std=True)),
('scaler' , MaxAbsScaler(copy=True)),
('classifier', SGDClassifier(verbose=0, class_weight='balanced',
random_state=randomSeeds['randForClassifier']) ),
] )
parameters={'vectorizer__ngram_range':[(1,2)],
'vectorizer__min_df':[2],
'vectorizer__max_df':[.98],
'classifier__alpha':[1],
'classifier__learning_rate':['invscaling'],
'classifier__eta0':[ .01],
'classifier__loss':[ 'hinge' ],
'classifier__penalty':['l2'],
}
p = tl.TextPipelineTuningHelper( pipeline, parameters,
trainingDataDir=args.trainingData,
testSplit=args.testSplit,
gridSearchBeta=args.gridSearchBeta,
gridSearchCV=args.gridSearchCV,
indexOfYes=args.indexOfYes,
randomSeeds=randomSeeds,
).fit()
print p.getReports(wIndex=args.wIndex,
tuningIndexFile=args.tuningIndexFile,
wPredictions=args.wPredictions,
predFilePrefix=args.predFilePrefix,
compareBeta=args.compareBeta,
verbose=args.verbose,
)
lowercase=False, # done in preprocessing
stop_words='english',
# token_pattern=r'\b([a-z_]\w+)\b', # use default for now
),
),
('scaler', StandardScaler(copy=True, with_mean=False, with_std=True)),
#('scaler' , MaxAbsScaler(copy=True)),
('classifier',
SGDClassifier(verbose=0, random_state=randomSeeds['randForClassifier'])),
])
parameters = {
'vectorizer__ngram_range': [(1, 2)],
'vectorizer__min_df': [0.1],
'vectorizer__max_df': [.7],
#'vectorizer__max_features':[2000],
'classifier__loss': ['hinge'],
'classifier__penalty': ['l2'],
'classifier__alpha': [5],
'classifier__learning_rate': ['optimal'],
'classifier__class_weight': ['balanced'],
'classifier__eta0': [.01],
}
p = tl.TextPipelineTuningHelper(
pipeline,
parameters,
beta=4,
cv=2,
randomSeeds=randomSeeds,
).fit()
print p.getReports()
verbose=1,
random_state=randomSeeds['randForClassifier'],
learning_rate=0.05,
n_estimators=1600,
max_depth=3,
min_samples_split=600,
min_samples_leaf=150,
max_features=0.7,
subsample=0.85,
)),
])
parameters = {
# 'classifier__learning_rate': [0.025],
# 'classifier__n_estimators': [3200,],
# 'classifier__max_depth': [3, 6, 9 ],
# 'classifier__min_samples_split': [525, 550, 575, 600, 625, ],
# 'classifier__min_samples_leaf': [100, 125, 150, 175],
# 'classifier__max_features': [0.6, 0.65, 0.7, 0.75, 0.8, 0.9, None, ],
# 'classifier__subsample': [0.6, 0.7, 0.75, 0.8, 0.85, 0.9, 0.95, 1.0,],
}
note = '\n'.join([
"blessed GB.",
]) + '\n'
p = tl.TextPipelineTuningHelper(
pipeline,
parameters,
randomSeeds=randomSeeds,
note=note,
).fit()
print(p.getReports())