})
rf_anova.fit(X_train_anova_p, y_train_p)
rf_anova_raw_preds, rf_anova_preds, rf_anova_score = rf_anova.predict(
X_test_anova_p, y_test_p)
rf_anova_fig = Classifiers.metrics(y_test_p, rf_anova_raw_preds[:, 1],
rf_anova_preds)
################################
######## GLM ROC Curves ########
glm_total = Classifiers.LogisticRegressionModel(params={
'penalty': 'l2',
'solver': 'lbfgs',
'max_iter': 200,
'C': 0.001,
'verbose': 0
})
glm_total.fit(X_train_p, y_train_p)
glm_total_raw_preds, glm_total_preds, glm_total_score = glm_total.predict(
X_test_p, y_test_p)
glm_total_fig = Classifiers.metrics(y_test_p, glm_total_raw_preds[:, 1],
glm_total_preds)
glm_vt = Classifiers.LogisticRegressionModel(params={
'penalty': 'l2',
'solver': 'lbfgs',
'max_iter': 200,
)
rf_anova.fit(X_train_p, y_train_p)
rf_anova_raw_preds, rf_anova_preds, rf_anova_score = rf_anova.predict(
X_test_p, y_test_p
)
auc_rf, pr_auc, fpr, tpr, roc_thresholds, recalls, precisions = get_auc_pr(
y_test_p, rf_anova_raw_preds
)
glm_anova = Classifiers.LogisticRegressionModel(
params={
**best_params_glm,
"penalty": "l2",
"solver": "lbfgs",
"max_iter": 200,
# 'C' : 0.001,
"verbose": 0,
}
)
glm_anova.fit(X_train_p, y_train_p)
glm_anova_raw_preds, glm_anova_preds, glm_anova_score = glm_anova.predict(
X_test_p, y_test_p
)
auc_glm, pr_auc_glm, fpr_glm, tpr_glm, roc_thresholds_glm, recalls_glm, precisions_glm = get_auc_pr(
y_test_p, glm_anova_raw_preds
)
features_used = [i[1] for i in nznew]
X_glm_train = X_train[features_used]
X_glm_test = X_test[features_used]
# Scale and normalize
X_train_new = preprocessing.scale(X_glm_train)
X_test_new = preprocessing.scale(X_glm_test)
print("Shape of X data GLM fr: ", X_train_new.shape)
# Use new features to train GLM model
glm = Classifiers.LogisticRegressionModel(params={
"penalty": "l2",
"solver": "lbfgs",
"max_iter": 200,
"C": 0.01,
"verbose": 0,
})
glm.fit(X_train_new, y_train_p)
# Save model to disk
filename = 'S:/Dehydration_stroke/Team Emerald/Working GitHub Directories/'\
'Michael/stroke-hemodynamics/Aim 2/Models/FullModelResults/UpdatedResults/TEST24hr_model_glm.sav'
pickle.dump(glm, open(filename, 'wb'))
glm_raw_preds, glm_preds, glm_score = glm.predict(X_test_new, y_test_p)
auc_glm, pr_auc_glm, fpr_glm, tpr_glm, roc_thresholds_glm, recalls_glm, precisions_glm = get_auc_pr(
y_test_p, glm_raw_preds)
recall_no_skill = y_train_p.sum() / len(y_train_p)
