From 40029a8205311d54f4e046325aba920df2a8b5dd Mon Sep 17 00:00:00 2001 From: Primoz Date: Mon, 21 Nov 2022 14:47:19 +0100 Subject: [PATCH] Add a script for ml classification pipeline. --- exploration/ml_pipeline_classification.py | 356 ++++++++++++++++++++++ exploration/ml_pipeline_regression.py | 4 +- 2 files changed, 358 insertions(+), 2 deletions(-) create mode 100644 exploration/ml_pipeline_classification.py diff --git a/exploration/ml_pipeline_classification.py b/exploration/ml_pipeline_classification.py new file mode 100644 index 0000000..140464b --- /dev/null +++ b/exploration/ml_pipeline_classification.py @@ -0,0 +1,356 @@ +# --- +# jupyter: +# jupytext: +# formats: ipynb,py:percent +# text_representation: +# extension: .py +# format_name: percent +# format_version: '1.3' +# jupytext_version: 1.13.0 +# kernelspec: +# display_name: straw2analysis +# language: python +# name: straw2analysis +# --- + +# %% jupyter={"source_hidden": true} +# %matplotlib inline +import datetime +import importlib +import os +import sys + +import numpy as np +import matplotlib.pyplot as plt +import pandas as pd +import seaborn as sns + +from sklearn import linear_model, svm, naive_bayes, neighbors, tree, ensemble +from sklearn.model_selection import LeaveOneGroupOut, cross_validate +from sklearn.dummy import DummyClassifier +from sklearn.impute import SimpleImputer + +from lightgbm import LGBMClassifier +import xgboost as xg +from IPython.core.interactiveshell import InteractiveShell +InteractiveShell.ast_node_interactivity = "all" + +nb_dir = os.path.split(os.getcwd())[0] +if nb_dir not in sys.path: + sys.path.append(nb_dir) + +import machine_learning.labels +import machine_learning.model + +# %% [markdown] +# # RAPIDS models + +# %% [markdown] +# ## PANAS negative affect + +# %% jupyter={"source_hidden": true} +model_input = pd.read_csv("../data/intradaily_30_min_all_targets/input_JCQ_job_demand_mean.csv") + +# %% jupyter={"source_hidden": true} +bins = [-4, -1, 1, 4] # bins for z-scored targets +model_input['target'], edges = pd.cut(model_input.target, bins=bins, labels=['low', 'medium', 'high'], retbins=True, right=False) +model_input['target'].value_counts(), edges +model_input = model_input[model_input['target'] != "medium"] +model_input['target'] = model_input['target'].astype(str).apply(lambda x: 0 if x == "low" else 1) + +model_input['target'].value_counts() + +# %% jupyter={"source_hidden": true} +index_columns = ["local_segment", "local_segment_label", "local_segment_start_datetime", "local_segment_end_datetime"] +#if "pid" in model_input.columns: +# index_columns.append("pid") +model_input.set_index(index_columns, inplace=True) + +# %% jupyter={"source_hidden": true} +cv_method = '5kfold' +if cv_method == 'logo': + data_x, data_y, data_groups = model_input.drop(["target", "pid"], axis=1), model_input["target"], model_input["pid"] +else: + model_input['pid_index'] = model_input.groupby('pid').cumcount() + model_input['pid_count'] = model_input.groupby('pid')['pid'].transform('count') + + model_input["pid_index"] = (model_input['pid_index'] / model_input['pid_count'] + 1).round() + model_input["pid_half"] = model_input["pid"] + "_" + model_input["pid_index"].astype(int).astype(str) + + data_x, data_y, data_groups = model_input.drop(["target", "pid", "pid_index", "pid_half"], axis=1), model_input["target"], model_input["pid_half"] + +# %% jupyter={"source_hidden": true} +categorical_feature_colnames = ["gender", "startlanguage"] +additional_categorical_features = [] #[col for col in data_x.columns if "mostcommonactivity" in col or "homelabel" in col] +categorical_feature_colnames += additional_categorical_features + +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) +train_x.dtypes + +# %% jupyter={"source_hidden": true} +logo = LeaveOneGroupOut() +logo.get_n_splits( + train_x, + data_y, + groups=data_groups, +) + +# Defaults to 5 k-folds in cross_validate method +if cv_method != 'logo' and cv_method != 'half_logo': + logo = None + +# %% jupyter={"source_hidden": true} +imputer = SimpleImputer(missing_values=np.nan, strategy='mean') + +# %% [markdown] +# ### Baseline: Dummy Classifier (most frequent) +dummy_class = DummyClassifier(strategy="most_frequent") + +# %% jupyter={"source_hidden": true} +dummy_classifier = cross_validate( + dummy_class, + X=imputer.fit_transform(train_x), + y=data_y, + groups=data_groups, + cv=logo, + n_jobs=-1, + scoring=('accuracy', 'average_precision', 'recall', 'f1') +) +# %% jupyter={"source_hidden": true} +print("Acc", np.median(dummy_classifier['test_accuracy'])) +print("Precision", np.median(dummy_classifier['test_average_precision'])) +print("Recall", np.median(dummy_classifier['test_recall'])) +print("F1", np.median(dummy_classifier['test_f1'])) + +# %% [markdown] +# ### Logistic Regression + +# %% jupyter={"source_hidden": true} +logistic_regression = linear_model.LogisticRegression() + +# %% jupyter={"source_hidden": true} +log_reg_scores = cross_validate( + logistic_regression, + X=imputer.fit_transform(train_x), + y=data_y, + groups=data_groups, + cv=logo, + n_jobs=-1, + scoring=('accuracy', 'precision', 'recall', 'f1') +) +# %% jupyter={"source_hidden": true} +print("Acc", np.median(log_reg_scores['test_accuracy'])) +print("Precision", np.median(log_reg_scores['test_precision'])) +print("Recall", np.median(log_reg_scores['test_recall'])) +print("F1", np.median(log_reg_scores['test_f1'])) + +# %% [markdown] +# ### Support Vector Machine + +# %% jupyter={"source_hidden": true} +svc = svm.SVC() + +# %% jupyter={"source_hidden": true} +svc_scores = cross_validate( + svc, + X=imputer.fit_transform(train_x), + y=data_y, + groups=data_groups, + cv=logo, + n_jobs=-1, + scoring=('accuracy', 'precision', 'recall', 'f1') +) +# %% jupyter={"source_hidden": true} +print("Acc", np.median(svc_scores['test_accuracy'])) +print("Precision", np.median(svc_scores['test_precision'])) +print("Recall", np.median(svc_scores['test_recall'])) +print("F1", np.median(svc_scores['test_f1'])) + +# %% [markdown] +# ### Gaussian Naive Bayes + +# %% jupyter={"source_hidden": true} +gaussian_nb = naive_bayes.GaussianNB() + +# %% jupyter={"source_hidden": true} +gaussian_nb_scores = cross_validate( + gaussian_nb, + X=imputer.fit_transform(train_x), + y=data_y, + groups=data_groups, + cv=logo, + n_jobs=-1, + scoring=('accuracy', 'precision', 'recall', 'f1') +) +# %% jupyter={"source_hidden": true} +print("Acc", np.median(gaussian_nb_scores['test_accuracy'])) +print("Precision", np.median(gaussian_nb_scores['test_precision'])) +print("Recall", np.median(gaussian_nb_scores['test_recall'])) +print("F1", np.median(gaussian_nb_scores['test_f1'])) + +# %% [markdown] +# ### Stochastic Gradient Descent Classifier + +# %% jupyter={"source_hidden": true} +sgdc = linear_model.SGDClassifier() + +# %% jupyter={"source_hidden": true} +sgdc_scores = cross_validate( + sgdc, + X=imputer.fit_transform(train_x), + y=data_y, + groups=data_groups, + cv=logo, + n_jobs=-1, + scoring=('accuracy', 'precision', 'recall', 'f1') +) +# %% jupyter={"source_hidden": true} +print("Acc", np.median(sgdc_scores['test_accuracy'])) +print("Precision", np.median(sgdc_scores['test_precision'])) +print("Recall", np.median(sgdc_scores['test_recall'])) +print("F1", np.median(sgdc_scores['test_f1'])) + +# %% [markdown] +# ### K-nearest neighbors + +# %% jupyter={"source_hidden": true} +knn = neighbors.KNeighborsClassifier() + +# %% jupyter={"source_hidden": true} +knn_scores = cross_validate( # Nekaj ne funkcionira pravilno + knn, + X=imputer.fit_transform(train_x), + y=data_y, + groups=data_groups, + cv=logo, + n_jobs=-1, + scoring=('accuracy', 'precision', 'recall', 'f1') + # error_score='raise' +) +# %% jupyter={"source_hidden": true} +print("Acc", np.median(knn_scores['test_accuracy'])) +print("Precision", np.median(knn_scores['test_precision'])) +print("Recall", np.median(knn_scores['test_recall'])) +print("F1", np.median(knn_scores['test_f1'])) + +# %% [markdown] +# ### Decision Tree + +# %% jupyter={"source_hidden": true} +dtree = tree.DecisionTreeClassifier() + +# %% jupyter={"source_hidden": true} +dtree_scores = cross_validate( + dtree, + X=imputer.fit_transform(train_x), + y=data_y, + groups=data_groups, + cv=logo, + n_jobs=-1, + scoring=('accuracy', 'precision', 'recall', 'f1') +) +# %% jupyter={"source_hidden": true} +print("Acc", np.median(dtree_scores['test_accuracy'])) +print("Precision", np.median(dtree_scores['test_precision'])) +print("Recall", np.median(dtree_scores['test_recall'])) +print("F1", np.median(dtree_scores['test_f1'])) + +# %% [markdown] +# ### Random Forest Classifier + +# %% jupyter={"source_hidden": true} +rfc = ensemble.RandomForestClassifier() + +# %% jupyter={"source_hidden": true} +rfc_scores = cross_validate( + rfc, + X=imputer.fit_transform(train_x), + y=data_y, + groups=data_groups, + cv=logo, + n_jobs=-1, + scoring=('accuracy', 'precision', 'recall', 'f1') +) +# %% jupyter={"source_hidden": true} +print("Acc", np.median(rfc_scores['test_accuracy'])) +print("Precision", np.median(rfc_scores['test_precision'])) +print("Recall", np.median(rfc_scores['test_recall'])) +print("F1", np.median(rfc_scores['test_f1'])) + +# %% [markdown] +# ### Gradient Boosting Classifier + +# %% jupyter={"source_hidden": true} +gbc = ensemble.GradientBoostingClassifier() + +# %% jupyter={"source_hidden": true} +gbc_scores = cross_validate( + gbc, + X=imputer.fit_transform(train_x), + y=data_y, + groups=data_groups, + cv=logo, + n_jobs=-1, + scoring=('accuracy', 'precision', 'recall', 'f1') +) +# %% jupyter={"source_hidden": true} +print("Acc", np.median(gbc_scores['test_accuracy'])) +print("Precision", np.median(gbc_scores['test_precision'])) +print("Recall", np.median(gbc_scores['test_recall'])) +print("F1", np.median(gbc_scores['test_f1'])) + +# %% [markdown] +# ### LGBM Classifier + +# %% jupyter={"source_hidden": true} +lgbm = LGBMClassifier() + +# %% jupyter={"source_hidden": true} +lgbm_scores = cross_validate( + lgbm, + X=imputer.fit_transform(train_x), + y=data_y, + groups=data_groups, + cv=logo, + n_jobs=-1, + scoring=('accuracy', 'precision', 'recall', 'f1') +) +# %% jupyter={"source_hidden": true} +print("Acc", np.median(lgbm_scores['test_accuracy'])) +print("Precision", np.median(lgbm_scores['test_precision'])) +print("Recall", np.median(lgbm_scores['test_recall'])) +print("F1", np.median(lgbm_scores['test_f1'])) + +# %% [markdown] +# ### XGBoost Classifier + +# %% jupyter={"source_hidden": true} +xgb_classifier = xg.sklearn.XGBClassifier() + +# %% jupyter={"source_hidden": true} +xgb_classifier_scores = cross_validate( + xgb_classifier, + X=imputer.fit_transform(train_x), + y=data_y, + groups=data_groups, + cv=logo, + n_jobs=-1, + scoring=('accuracy', 'precision', 'recall', 'f1') +) +# %% jupyter={"source_hidden": true} +print("Acc", np.median(xgb_classifier_scores['test_accuracy'])) +print("Precision", np.median(xgb_classifier_scores['test_precision'])) +print("Recall", np.median(xgb_classifier_scores['test_recall'])) +print("F1", np.median(xgb_classifier_scores['test_f1'])) diff --git a/exploration/ml_pipeline_regression.py b/exploration/ml_pipeline_regression.py index 21d02cb..98b2e3f 100644 --- a/exploration/ml_pipeline_regression.py +++ b/exploration/ml_pipeline_regression.py @@ -58,8 +58,8 @@ index_columns = ["local_segment", "local_segment_label", "local_segment_start_da # index_columns.append("pid") model_input.set_index(index_columns, inplace=True) -cv_method = '5kfold' -if cv_method == 'half_logo': +cv_method = 'half_logo' # logo, half_logo, 5kfold +if cv_method == 'logo': data_x, data_y, data_groups = model_input.drop(["target", "pid"], axis=1), model_input["target"], model_input["pid"] else: model_input['pid_index'] = model_input.groupby('pid').cumcount()