log = logging.getLogger()
handler = logging.StreamHandler()
formatter = logging.Formatter(
'%(asctime)s %(name)-12s %(levelname)-8s %(message)s')
handler.setFormatter(formatter)
log.addHandler(handler)
log.setLevel(logging.DEBUG)
'''
Pipeline example using scikit-learn
- Scaling features
- Training different models and hyperparameter sets
- Feature selection using SelectPercentile
'''
pip = Pipeline(config, load_yaml('exp.yaml'), workers=10, save=True)
@pip.load
def load(config):
data = {}
# data loading
train = pd.read_csv('../data/train.csv', index_col='id')
data['train_x'] = train.drop('survived', axis=1).values
data['train_y'] = train.survived.values
test = pd.read_csv('../data/test.csv', index_col='id')
data['test_x'] = test.values
data['test_ids'] = test.index
return data
from dstools.util import load_yaml
from dstools.lab import Experiment
import pandas as pd
ex = Experiment(load_yaml('exp.yaml')['conf'])
ex.get(_id=['57435574e0f48c88752991ba'])
best = ex.records[0]
df = pd.DataFrame(best.test_preds, columns=['PassengerId', 'Survived'])
df.set_index('PassengerId', inplace=True)
df.Survived = df.Survived.astype(int)
df.to_csv('res.csv')
from dstools.util import load_yaml
from dstools.sklearn import grid_generator
from pipeline.lab.util import top_k
from pipeline import SKPipeline
# logger configuration
log = logging.getLogger()
handler = logging.StreamHandler()
formatter = logging.Formatter(
'%(asctime)s %(name)-12s %(levelname)-8s %(message)s')
handler.setFormatter(formatter)
log.addHandler(handler)
log.setLevel(logging.INFO)
# create pipeline object
pip = SKPipeline(config, load_yaml('exp.yaml'))
# this function should return the all the data used to train models
# as a dictionary. In subsequentent functions the data will be available in
# the 'data' parameter
@pip.load
def load(config):
iris = load_iris()
X_train, X_test, y_train, y_test = train_test_split(iris.data,
iris.target,
test_size=0.30,
random_state=0)
data = {
'X_train': X_train,
'X_test': X_test,
record['mean_acc'] = scores.mean()
# predict on the test set
fn = SelectPercentile(f_classif, percentile).fit(train_x, train_y)
train_x = fn.transform(train_x)
test_x = fn.transform(test_x)
scaler = StandardScaler().fit(train_x)
train_x = scaler.transform(train_x)
test_x = scaler.transform(test_x)
model.fit(train_x, train_y)
ids = data['test_ids']
preds = model.predict(test_x)
record['test_preds'] = [(id_, pred) for id_, pred in zip(ids, preds)]
def finalize(config, experiment):
experiment.records = top_k(experiment.records, 'mean_acc', 30)
experiment['exp_name'] = config['exp_name']
pip = Pipeline(config, load_yaml('exp.yaml'), workers=4, save=True)
pip.load = load
# pip.model_iterator = model_iterator
pip.model_iterator = model_iterator_autosklearn
pip.train = train
pip.finalize = finalize
pip()
# function used to train models, should return
# a fitted model
def train(config, model, data, record):
print record
model.fit(data['x_train'], data['y_train'])
preds = model.predict(data['X_test'])
record['precision'] = precision_score(data['y_test'], preds)
return model
# optional function used when every model has been trained
def finalize(config, experiment):
pass
# experiment.records = top_k(experiment.records, 'precision', 4)
# create pipeline object
pip = SKPipeline(config, load_yaml('exp.yaml'), workers=1)
# assign your functions
pip.load = load
pip.model_iterator = model_iterator
pip.train = train
pip.finalize = finalize
# run pipeline
pip()