def predict(self, dataset):
log.debug("getting features...")
X_test = tools.get_features(dataset)
log.debug("predicting with model...")
self.preds = self.model.predict(X_test)
return self.preds
def transform(self, dataset, drop_categorical=True, return_df=False):
"""
extracts features from dataset and returns as numpy array
:param dataset: dataset to extract features from
:param drop_categorical: if true, drop categorical features
:param return_df: if true, return result as pandas dataframe, else as numpy array of values
:return: numpy array with features for each sample in dataset
"""
return tools.get_features(dataset, drop_categorical=drop_categorical, return_df=return_df)
def predict(self, dataset):
log.debug("getting features...")
df_test, categorical = tools.get_features(dataset, return_df=True, drop_categorical=False, return_categorical_list=True)
self.d_test = lgb.Dataset(df_test,
label=dataset.labels,
categorical_feature=categorical)
log.debug("predicting with model...")
self.preds = self.model.predict(df_test)
return self.preds
def fit(self, dataset, ROUNDS=100):
log.debug("getting features...")
df_train, categorical = tools.get_features(dataset, return_df=True, drop_categorical=False, return_categorical_list=True)
self.d_train = lgb.Dataset(df_train, label=dataset.labels, categorical_feature=categorical)
params = {
'task': 'train',
'boosting_type': 'gbdt',
'objective': 'binary',
'metric': {'binary_logloss'},
'num_leaves': 96,
'max_depth': 10,
'feature_fraction': 0.9,
'bagging_fraction': 0.95,
'bagging_freq': 5
}
log.debug("fitting classifier...")
self.model = lgb.train(params, self.d_train, ROUNDS)
def fit(self, dataset):
log.debug("getting features...")
X_train = tools.get_features(dataset)
log.debug("fitting classifier...")
self.model.fit(X_train, dataset.labels)