2021-09-15 15:14:54 +02:00
|
|
|
from pathlib import Path
|
2022-11-16 18:13:30 +01:00
|
|
|
from sklearn import linear_model, svm, kernel_ridge, gaussian_process, ensemble
|
2022-12-07 21:25:05 +01:00
|
|
|
from sklearn.model_selection import LeaveOneGroupOut, cross_validate, cross_validate
|
2022-11-16 18:13:30 +01:00
|
|
|
from sklearn.metrics import mean_squared_error, r2_score
|
|
|
|
from sklearn.impute import SimpleImputer
|
|
|
|
from sklearn.dummy import DummyRegressor
|
|
|
|
from xgboost import XGBRegressor
|
2021-09-15 15:14:54 +02:00
|
|
|
|
|
|
|
import pandas as pd
|
2022-11-16 18:13:30 +01:00
|
|
|
import numpy as np
|
2021-09-15 15:14:54 +02:00
|
|
|
|
|
|
|
|
|
|
|
def safe_outer_merge_on_index(left: pd.DataFrame, right: pd.DataFrame) -> pd.DataFrame:
|
|
|
|
if left.empty:
|
|
|
|
return right
|
|
|
|
elif right.empty:
|
|
|
|
return left
|
|
|
|
else:
|
|
|
|
return pd.merge(
|
|
|
|
left,
|
|
|
|
right,
|
|
|
|
how="outer",
|
|
|
|
left_index=True,
|
|
|
|
right_index=True,
|
|
|
|
validate="one_to_one",
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
def to_csv_with_settings(
|
|
|
|
df: pd.DataFrame, folder: Path, filename_prefix: str, data_type: str
|
|
|
|
) -> None:
|
|
|
|
full_path = construct_full_path(folder, filename_prefix, data_type)
|
|
|
|
df.to_csv(
|
|
|
|
path_or_buf=full_path,
|
|
|
|
sep=",",
|
|
|
|
na_rep="NA",
|
|
|
|
header=True,
|
|
|
|
index=True,
|
|
|
|
encoding="utf-8",
|
|
|
|
)
|
|
|
|
print("Exported the dataframe to " + str(full_path))
|
|
|
|
|
|
|
|
|
|
|
|
def read_csv_with_settings(
|
|
|
|
folder: Path, filename_prefix: str, data_type: str, grouping_variable: list
|
|
|
|
) -> pd.DataFrame:
|
|
|
|
full_path = construct_full_path(folder, filename_prefix, data_type)
|
|
|
|
return pd.read_csv(
|
|
|
|
filepath_or_buffer=full_path,
|
|
|
|
sep=",",
|
|
|
|
header=0,
|
|
|
|
na_values="NA",
|
|
|
|
encoding="utf-8",
|
|
|
|
index_col=(["participant_id"] + grouping_variable),
|
|
|
|
parse_dates=True,
|
|
|
|
infer_datetime_format=True,
|
|
|
|
cache_dates=True,
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
def construct_full_path(folder: Path, filename_prefix: str, data_type: str) -> Path:
|
|
|
|
export_filename = filename_prefix + "_" + data_type + ".csv"
|
|
|
|
full_path = folder / export_filename
|
|
|
|
return full_path
|
2022-11-16 18:13:30 +01:00
|
|
|
|
|
|
|
def insert_row(df, row):
|
|
|
|
return pd.concat([df, pd.DataFrame([row], columns=df.columns)], ignore_index=True)
|
|
|
|
|
2022-12-07 21:25:05 +01:00
|
|
|
def prepare_regression_model_input(input_csv):
|
2022-11-16 19:34:18 +01:00
|
|
|
|
2022-11-16 18:13:30 +01:00
|
|
|
model_input = pd.read_csv(input_csv)
|
|
|
|
|
|
|
|
index_columns = ["local_segment", "local_segment_label", "local_segment_start_datetime", "local_segment_end_datetime"]
|
|
|
|
model_input.set_index(index_columns, inplace=True)
|
|
|
|
|
|
|
|
data_x, data_y, data_groups = model_input.drop(["target", "pid"], axis=1), model_input["target"], model_input["pid"]
|
|
|
|
|
2022-11-16 19:34:18 +01:00
|
|
|
categorical_feature_colnames = ["gender", "startlanguage", "limesurvey_demand_control_ratio_quartile"]
|
2022-12-07 21:25:05 +01:00
|
|
|
additional_categorical_features = [col for col in data_x.columns if "mostcommonactivity" in col or "homelabel" in col]
|
|
|
|
categorical_feature_colnames += additional_categorical_features
|
2022-11-16 19:34:18 +01:00
|
|
|
#TODO: check whether limesurvey_demand_control_ratio_quartile NaNs could be replaced meaningfully
|
2022-11-16 18:13:30 +01:00
|
|
|
categorical_features = data_x[categorical_feature_colnames].copy()
|
|
|
|
mode_categorical_features = categorical_features.mode().iloc[0]
|
|
|
|
# fillna with mode
|
|
|
|
categorical_features = categorical_features.fillna(mode_categorical_features)
|
|
|
|
# one-hot encoding
|
|
|
|
categorical_features = categorical_features.apply(lambda col: col.astype("category"))
|
|
|
|
if not categorical_features.empty:
|
|
|
|
categorical_features = pd.get_dummies(categorical_features)
|
|
|
|
|
|
|
|
numerical_features = data_x.drop(categorical_feature_colnames, axis=1)
|
|
|
|
|
|
|
|
train_x = pd.concat([numerical_features, categorical_features], axis=1)
|
|
|
|
|
2022-11-16 19:34:18 +01:00
|
|
|
return train_x, data_y, data_groups
|
|
|
|
|
|
|
|
|
2022-12-07 21:25:05 +01:00
|
|
|
def run_all_regression_models(input_csv):
|
2022-11-16 19:34:18 +01:00
|
|
|
# Prepare data
|
2022-12-07 21:25:05 +01:00
|
|
|
data_x, data_y, data_groups = prepare_regression_model_input(input_csv)
|
2022-11-16 19:34:18 +01:00
|
|
|
|
2022-11-16 18:13:30 +01:00
|
|
|
# Prepare cross validation
|
|
|
|
logo = LeaveOneGroupOut()
|
|
|
|
logo.get_n_splits(
|
2022-12-07 21:25:05 +01:00
|
|
|
data_x,
|
2022-11-16 18:13:30 +01:00
|
|
|
data_y,
|
|
|
|
groups=data_groups,
|
|
|
|
)
|
2022-12-07 21:25:05 +01:00
|
|
|
metrics = ['r2', 'neg_mean_absolute_error', 'neg_root_mean_squared_error']
|
|
|
|
test_metrics = ["test_" + metric for metric in metrics]
|
|
|
|
scores = pd.DataFrame(columns=["method", "max", "nanmedian"])
|
2022-11-16 18:13:30 +01:00
|
|
|
|
|
|
|
# Validate models
|
2022-12-07 21:25:05 +01:00
|
|
|
dummy_regr = DummyRegressor(strategy="mean")
|
|
|
|
dummy_regr_scores = cross_validate(
|
|
|
|
dummy_regr,
|
|
|
|
X=data_x,
|
|
|
|
y=data_y,
|
|
|
|
groups=data_groups,
|
|
|
|
cv=logo,
|
|
|
|
n_jobs=-1,
|
|
|
|
scoring=metrics
|
|
|
|
)
|
|
|
|
print("Dummy model:")
|
|
|
|
print("R^2: ", np.nanmedian(dummy_regr_scores['test_r2']))
|
|
|
|
|
|
|
|
scores_df = pd.DataFrame(dummy_regr_scores)[test_metrics]
|
|
|
|
scores_df = scores_df.agg(['max', np.nanmedian]).transpose()
|
|
|
|
scores_df["method"] = "dummy"
|
|
|
|
scores = pd.concat([scores, scores_df])
|
|
|
|
|
2022-11-16 18:13:30 +01:00
|
|
|
lin_reg_rapids = linear_model.LinearRegression()
|
2022-12-07 21:25:05 +01:00
|
|
|
lin_reg_scores = cross_validate(
|
2022-11-16 18:13:30 +01:00
|
|
|
lin_reg_rapids,
|
2022-12-07 21:25:05 +01:00
|
|
|
X=data_x,
|
2022-11-16 18:13:30 +01:00
|
|
|
y=data_y,
|
|
|
|
groups=data_groups,
|
|
|
|
cv=logo,
|
|
|
|
n_jobs=-1,
|
2022-12-07 21:25:05 +01:00
|
|
|
scoring=metrics
|
2022-11-16 18:13:30 +01:00
|
|
|
)
|
|
|
|
print("Linear regression:")
|
2022-12-07 21:25:05 +01:00
|
|
|
print("R^2: ", np.nanmedian(lin_reg_scores['test_r2']))
|
|
|
|
|
|
|
|
scores_df = pd.DataFrame(lin_reg_scores)[test_metrics]
|
|
|
|
scores_df = scores_df.agg(['max', np.nanmedian]).transpose()
|
|
|
|
scores_df["method"] = "linear_reg"
|
|
|
|
scores = pd.concat([scores, scores_df])
|
2022-11-16 18:13:30 +01:00
|
|
|
|
|
|
|
ridge_reg = linear_model.Ridge(alpha=.5)
|
2022-12-07 21:25:05 +01:00
|
|
|
ridge_reg_scores = cross_validate(
|
2022-11-16 18:13:30 +01:00
|
|
|
ridge_reg,
|
2022-12-07 21:25:05 +01:00
|
|
|
X=data_x,
|
2022-11-16 18:13:30 +01:00
|
|
|
y=data_y,
|
|
|
|
groups=data_groups,
|
|
|
|
cv=logo,
|
|
|
|
n_jobs=-1,
|
2022-12-07 21:25:05 +01:00
|
|
|
scoring=metrics
|
2022-11-16 18:13:30 +01:00
|
|
|
)
|
|
|
|
print("Ridge regression")
|
2022-12-07 21:25:05 +01:00
|
|
|
|
|
|
|
scores_df = pd.DataFrame(ridge_reg_scores)[test_metrics]
|
|
|
|
scores_df = scores_df.agg(['max', np.nanmedian]).transpose()
|
|
|
|
scores_df["method"] = "ridge_reg"
|
|
|
|
scores = pd.concat([scores, scores_df])
|
|
|
|
|
2022-11-16 18:13:30 +01:00
|
|
|
|
|
|
|
lasso_reg = linear_model.Lasso(alpha=0.1)
|
2022-12-07 21:25:05 +01:00
|
|
|
lasso_reg_score = cross_validate(
|
2022-11-16 18:13:30 +01:00
|
|
|
lasso_reg,
|
2022-12-07 21:25:05 +01:00
|
|
|
X=data_x,
|
2022-11-16 18:13:30 +01:00
|
|
|
y=data_y,
|
|
|
|
groups=data_groups,
|
|
|
|
cv=logo,
|
|
|
|
n_jobs=-1,
|
2022-12-07 21:25:05 +01:00
|
|
|
scoring=metrics
|
2022-11-16 18:13:30 +01:00
|
|
|
)
|
|
|
|
print("Lasso regression")
|
2022-12-07 21:25:05 +01:00
|
|
|
|
|
|
|
scores_df = pd.DataFrame(lasso_reg_score)[test_metrics]
|
|
|
|
scores_df = scores_df.agg(['max', np.nanmedian]).transpose()
|
|
|
|
scores_df["method"] = "lasso_reg"
|
|
|
|
scores = pd.concat([scores, scores_df])
|
2022-11-16 18:13:30 +01:00
|
|
|
|
|
|
|
bayesian_ridge_reg = linear_model.BayesianRidge()
|
2022-12-07 21:25:05 +01:00
|
|
|
bayesian_ridge_reg_score = cross_validate(
|
2022-11-16 18:13:30 +01:00
|
|
|
bayesian_ridge_reg,
|
2022-12-07 21:25:05 +01:00
|
|
|
X=data_x,
|
2022-11-16 18:13:30 +01:00
|
|
|
y=data_y,
|
|
|
|
groups=data_groups,
|
|
|
|
cv=logo,
|
|
|
|
n_jobs=-1,
|
2022-12-07 21:25:05 +01:00
|
|
|
scoring=metrics
|
2022-11-16 18:13:30 +01:00
|
|
|
)
|
|
|
|
print("Bayesian Ridge")
|
2022-12-07 21:25:05 +01:00
|
|
|
|
|
|
|
scores_df = pd.DataFrame(bayesian_ridge_reg_score)[test_metrics]
|
|
|
|
scores_df = scores_df.agg(['max', np.nanmedian]).transpose()
|
|
|
|
scores_df["method"] = "bayesian_ridge"
|
|
|
|
scores = pd.concat([scores, scores_df])
|
2022-11-16 18:13:30 +01:00
|
|
|
|
|
|
|
ransac_reg = linear_model.RANSACRegressor()
|
2022-12-07 21:25:05 +01:00
|
|
|
ransac_reg_score = cross_validate(
|
2022-11-16 18:13:30 +01:00
|
|
|
ransac_reg,
|
2022-12-07 21:25:05 +01:00
|
|
|
X=data_x,
|
2022-11-16 18:13:30 +01:00
|
|
|
y=data_y,
|
|
|
|
groups=data_groups,
|
|
|
|
cv=logo,
|
|
|
|
n_jobs=-1,
|
2022-12-07 21:25:05 +01:00
|
|
|
scoring=metrics
|
2022-11-16 18:13:30 +01:00
|
|
|
)
|
|
|
|
print("RANSAC (outlier robust regression)")
|
2022-12-07 21:25:05 +01:00
|
|
|
|
|
|
|
scores_df = pd.DataFrame(ransac_reg_score)[test_metrics]
|
|
|
|
scores_df = scores_df.agg(['max', np.nanmedian]).transpose()
|
|
|
|
scores_df["method"] = "RANSAC"
|
|
|
|
scores = pd.concat([scores, scores_df])
|
2022-11-16 18:13:30 +01:00
|
|
|
|
|
|
|
svr = svm.SVR()
|
2022-12-07 21:25:05 +01:00
|
|
|
svr_score = cross_validate(
|
2022-11-16 18:13:30 +01:00
|
|
|
svr,
|
2022-12-07 21:25:05 +01:00
|
|
|
X=data_x,
|
2022-11-16 18:13:30 +01:00
|
|
|
y=data_y,
|
|
|
|
groups=data_groups,
|
|
|
|
cv=logo,
|
|
|
|
n_jobs=-1,
|
2022-12-07 21:25:05 +01:00
|
|
|
scoring=metrics
|
2022-11-16 18:13:30 +01:00
|
|
|
)
|
|
|
|
print("Support vector regression")
|
2022-12-07 21:25:05 +01:00
|
|
|
|
|
|
|
scores_df = pd.DataFrame(svr_score)[test_metrics]
|
|
|
|
scores_df = scores_df.agg(['max', np.nanmedian]).transpose()
|
|
|
|
scores_df["method"] = "SVR"
|
|
|
|
scores = pd.concat([scores, scores_df])
|
2022-11-16 18:13:30 +01:00
|
|
|
|
|
|
|
kridge = kernel_ridge.KernelRidge()
|
2022-12-07 21:25:05 +01:00
|
|
|
kridge_score = cross_validate(
|
2022-11-16 18:13:30 +01:00
|
|
|
kridge,
|
2022-12-07 21:25:05 +01:00
|
|
|
X=data_x,
|
2022-11-16 18:13:30 +01:00
|
|
|
y=data_y,
|
|
|
|
groups=data_groups,
|
|
|
|
cv=logo,
|
|
|
|
n_jobs=-1,
|
2022-12-07 21:25:05 +01:00
|
|
|
scoring=metrics
|
2022-11-16 18:13:30 +01:00
|
|
|
)
|
|
|
|
print("Kernel Ridge regression")
|
2022-12-07 21:25:05 +01:00
|
|
|
|
|
|
|
scores_df = pd.DataFrame(kridge_score)[test_metrics]
|
|
|
|
scores_df = scores_df.agg(['max', np.nanmedian]).transpose()
|
|
|
|
scores_df["method"] = "kernel_ridge"
|
|
|
|
scores = pd.concat([scores, scores_df])
|
2022-11-16 18:13:30 +01:00
|
|
|
|
|
|
|
gpr = gaussian_process.GaussianProcessRegressor()
|
2022-12-07 21:25:05 +01:00
|
|
|
gpr_score = cross_validate(
|
2022-11-16 18:13:30 +01:00
|
|
|
gpr,
|
2022-12-07 21:25:05 +01:00
|
|
|
X=data_x,
|
2022-11-16 18:13:30 +01:00
|
|
|
y=data_y,
|
|
|
|
groups=data_groups,
|
|
|
|
cv=logo,
|
|
|
|
n_jobs=-1,
|
2022-12-07 21:25:05 +01:00
|
|
|
scoring=metrics
|
2022-11-16 18:13:30 +01:00
|
|
|
)
|
|
|
|
print("Gaussian Process Regression")
|
2022-12-07 21:25:05 +01:00
|
|
|
|
|
|
|
scores_df = pd.DataFrame(gpr_score)[test_metrics]
|
|
|
|
scores_df = scores_df.agg(['max', np.nanmedian]).transpose()
|
|
|
|
scores_df["method"] = "gaussian_proc"
|
|
|
|
scores = pd.concat([scores, scores_df])
|
2022-11-16 18:13:30 +01:00
|
|
|
|
|
|
|
rfr = ensemble.RandomForestRegressor(max_features=0.3, n_jobs=-1)
|
2022-12-07 21:25:05 +01:00
|
|
|
rfr_score = cross_validate(
|
2022-11-16 18:13:30 +01:00
|
|
|
rfr,
|
2022-12-07 21:25:05 +01:00
|
|
|
X=data_x,
|
2022-11-16 18:13:30 +01:00
|
|
|
y=data_y,
|
|
|
|
groups=data_groups,
|
|
|
|
cv=logo,
|
|
|
|
n_jobs=-1,
|
2022-12-07 21:25:05 +01:00
|
|
|
scoring=metrics
|
2022-11-16 18:13:30 +01:00
|
|
|
)
|
|
|
|
print("Random Forest Regression")
|
2022-12-07 21:25:05 +01:00
|
|
|
|
|
|
|
scores_df = pd.DataFrame(rfr_score)[test_metrics]
|
|
|
|
scores_df = scores_df.agg(['max', np.nanmedian]).transpose()
|
|
|
|
scores_df["method"] = "random_forest"
|
|
|
|
scores = pd.concat([scores, scores_df])
|
2022-11-16 18:13:30 +01:00
|
|
|
|
|
|
|
xgb = XGBRegressor()
|
2022-12-07 21:25:05 +01:00
|
|
|
xgb_score = cross_validate(
|
2022-11-16 18:13:30 +01:00
|
|
|
xgb,
|
2022-12-07 21:25:05 +01:00
|
|
|
X=data_x,
|
2022-11-16 18:13:30 +01:00
|
|
|
y=data_y,
|
|
|
|
groups=data_groups,
|
|
|
|
cv=logo,
|
|
|
|
n_jobs=-1,
|
2022-12-07 21:25:05 +01:00
|
|
|
scoring=metrics
|
2022-11-16 18:13:30 +01:00
|
|
|
)
|
|
|
|
print("XGBoost Regressor")
|
2022-12-07 21:25:05 +01:00
|
|
|
|
|
|
|
scores_df = pd.DataFrame(xgb_score)[test_metrics]
|
|
|
|
scores_df = scores_df.agg(['max', np.nanmedian]).transpose()
|
|
|
|
scores_df["method"] = "XGBoost"
|
|
|
|
scores = pd.concat([scores, scores_df])
|
2022-11-16 18:13:30 +01:00
|
|
|
|
|
|
|
ada = ensemble.AdaBoostRegressor()
|
2022-12-07 21:25:05 +01:00
|
|
|
ada_score = cross_validate(
|
2022-11-16 18:13:30 +01:00
|
|
|
ada,
|
2022-12-07 21:25:05 +01:00
|
|
|
X=data_x,
|
2022-11-16 18:13:30 +01:00
|
|
|
y=data_y,
|
|
|
|
groups=data_groups,
|
|
|
|
cv=logo,
|
|
|
|
n_jobs=-1,
|
2022-12-07 21:25:05 +01:00
|
|
|
scoring=metrics
|
2022-11-16 18:13:30 +01:00
|
|
|
)
|
|
|
|
print("ADA Boost Regressor")
|
2022-12-07 21:25:05 +01:00
|
|
|
|
|
|
|
scores_df = pd.DataFrame(ada_score)[test_metrics]
|
|
|
|
scores_df = scores_df.agg(['max', np.nanmedian]).transpose()
|
|
|
|
scores_df["method"] = "ADA_boost"
|
|
|
|
scores = pd.concat([scores, scores_df])
|
2022-11-16 18:13:30 +01:00
|
|
|
|
|
|
|
return scores
|