def train_and_persist(model_dir=None, hour_path=None, model="xgboost"):
hour = read_data(hour_path)
hour = preprocess(hour)
hour = dummify(hour)
hour = postprocess(hour)
train_X, test_X, train_y, test_y = split_train_test(hour)
# TODO: Implement other models?
if model == "xgboost":
model = train_xgboost(hour)
model_path = get_model_path(model_dir)
train_score = model.score(test_X, test_y)
joblib.dump(model, model_path + "/xgboost.pkl")
return train_score
elif model == "ridge":
model = train_ridge(hour)
model_path = get_model_path(model_dir)
train_score = model.score(test_X, test_y)
joblib.dump(model, model_path + "/ridge.pkl")
return train_score
elif model == "lasso":
model = train_lasso(hour)
model_path = get_model_path(model_dir)
train_score = model.score(test_X, test_y)
joblib.dump(model, model_path + "/lasso.pkl")
return train_score
def score(model_dir=None, hour_path=None, model_name="xgboost"):
model_path = get_model_path(model_dir, model=model_name)
if not os.path.exists(model_path):
train_and_persist(model_dir, model=model_name)
model_clf = joblib.load(model_path)
hour = read_data(hour_path)
hour = preprocess(hour)
hour = dummify(hour)
hour = postprocess(hour)
hour_d = pd.get_dummies(hour)
regex = re.compile(r"\[|\]|<", re.IGNORECASE)
hour_d.columns = [
regex.sub("_", col) if any(x in str(col)
for x in set(("[", "]", "<"))) else col
for col in hour_d.columns.values
]
hour_d = hour_d.select_dtypes(exclude="category")
hour_d_train_x, hour_d_test_x, hour_d_train_y, hour_d_test_y = split_train_test(
hour_d)
r2_train = model_clf.score(hour_d_train_x, hour_d_train_y)
r2_test = model_clf.score(hour_d_test_x, hour_d_test_y)
return r2_train, r2_test
def predict(parameters, model_dir=None, model=None):
"""Returns model prediction.
"""
if model == "ridge":
model_path = get_model_path(model_dir, model_al="ridge")
train_and_persist(model_dir=model_dir, model="ridge")
else:
model_path = get_model_path(model_dir, model_al="xgboost")
train_and_persist(model_dir=model_dir, model="xgboost")
model = joblib.load(model_path)
input_dict = get_input_dict(parameters)
X_input = pd.DataFrame([pd.Series(input_dict)])
result = model.predict(X_input)
# Undo np.sqrt(hour["cnt"])
return int(result ** 2)
def train_and_persist(model_dir=None, hour_path=None):
hour = read_data(hour_path)
hour = preprocess(hour)
hour = dummify(hour)
hour = postprocess(hour)
# TODO: Implement other models?
model = train_xgboost(hour)
model_path = get_model_path(model_dir)
joblib.dump(model, model_path)
def train_and_persist(model_dir=None, hour_path=None, model="xgboost"):
hour = read_data(hour_path)
hour = preprocess(hour)
hour = dummify(hour)
hour = postprocess(hour)
# todo:Implement other models?
if model == "xgboost":
train_model = train_xgboost(hour)
elif model == "ridge":
train_model = train_ridge(hour)
model_path = get_model_path(model_dir, model)
joblib.dump(train_model, model_path)
def predict(parameters, model_dir=None, model="xgboost"):
"""Returns model prediction.
"""
model_path = get_model_path(model_dir)
if not os.path.exists(model_path):
train_and_persist(model_dir=model_dir, model=model)
model = joblib.load(model_path)
input_dict = get_input_dict(parameters)
X_input = pd.DataFrame([pd.Series(input_dict)])
result = model.predict(X_input)
# Undo np.sqrt(hour["cnt"])
return int(result**2)
def train_and_persist(model_dir=None, hour_path=None, model="xgboost"):
"""
Trains and persists the model (xgboost by default, the other option is ridge).
"""
hour = read_data(hour_path)
hour = preprocess(hour)
hour = dummify(hour)
hour = postprocess(hour)
if model == "ridge":
model_clf = train_ridge(hour)
else:
model_clf = train_xgboost(hour)
model_path = get_model_path(model_dir, model)
joblib.dump(model_clf, model_path)
def train_and_persist(model_dir=None, hour_path=None, model="xgboost"):
hour = read_data(hour_path)
hour = preprocess(hour)
hour = dummify(hour)
hour = postprocess(hour)
model_path = get_model_path(model_dir, model)
if model == "xgboost":
scikit_model, score = train_xgboost(hour)
elif model == "ridge":
scikit_model, score = train_ridge(hour)
joblib.dump(scikit_model, model_path)
return score
def train_and_persist(model_dir=None, hour_path=None, model="xgboost"):
valid = {"xgboost", "ridge"}
if model not in valid:
raise ValueError("results: model must be one of %s." % valid)
else:
hour = read_data(hour_path)
hour = preprocess(hour)
hour = dummify(hour)
hour = postprocess(hour)
if model == "xgboost":
model_result = train_xgboost(hour)
else:
model_result = train_ridge(hour)
model_path = get_model_path(model, model_dir)
joblib.dump(model_result, model_path)
def train_and_persist(model_dir=None, hour_path=None, model="xgboost"):
hour = read_data(hour_path)
hour = preprocess(hour)
hour = dummify(hour)
hour = postprocess(hour)
model_path = get_model_path(model_dir, model)
# Implement other models?
if model == "ridge":
model_object, score_train = train_ridge(hour)
else:
model_object, score_train = train_xgboost(hour)
joblib.dump(model_object, model_path)
return score_train
def train_and_persist(model_dir=None, hour_path=None, model_name="xgboost"):
hour = read_data(hour_path)
hour = preprocess(hour)
hour = dummify(hour)
hour = postprocess(hour)
##added
# TODO: Implement other models?
if model_name == "xgboost":
model = train_xgboost(hour)
elif model_name == "ridge":
model = train_ridge(hour)
elif model_name == "lasso":
model = train_lasso(hour)
else:
print("model should be equal to 'xgboost' or 'ridge' or 'lasso'")
model_path = get_model_path(model_dir, model_name)
joblib.dump(model, model_path)