def run_random_forest_pipeline():
###### DATA LOADING
xy = XYLOADER() # CAN CHANGE
X = xy["X_train"]
y = xy["y_train"]
X_val = xy["X_val"]
y_val = xy["y_val"]
X_test = xy["X_test"]
output_index = xy["X_test_index"]
print("LOADED DATA")
###### PIPELINE/CV VARIABLES
###### DO NOT CHANGE BEFORE
clf = RandomForestClassifier()
fl = X.shape[1] # use for n_components
cv_grid = {
"clf__n_estimators": [100],
"clf__criterion": ["gini"],
"clf__min_samples_split": np.linspace(1, 20, 3).astype(int),
"clf__min_samples_leaf": np.linspace(1, 20, 3).astype(int),
}
num_folds = 3
####### START PREDICTIONS
print("TRAINING ESTIMATOR")
pred_pipe = Pipeline(steps=[("clf", clf)])
###### DO NOT CHANGE AFTER
estimator = GridSearchCV(pred_pipe, cv_grid, cv=num_folds)
# DO NOT NEED TO CHANGE BEYOND THIS LINE
KH.start_pipeline()
KH.record_metric("validation", "start", "prepross", preprocess_called, "", "")
KH.record_metric("validation", "start", estimator, "training", "", "")
estimator.fit(X, y)
KH.record_metric("validation", "end", estimator, "training", "", "")
KH.record_metric("validation", "end", estimator, "best_params", str(estimator.best_params_), "NA")
KH.record_metric("validation", "end", estimator, "best_estimator", str(estimator.best_estimator_), "NA")
KH.record_metric("validation", "end", estimator, "best_score", str(estimator.best_score_), "NA")
validation_score = str(estimator.score(X_val, y_val))
KH.record_metric("validation", "end", estimator, "validation score", validation_score, "")
preds = estimator.predict(X_test)
predictions = pd.DataFrame({"VisitNumber": output_index, "TripType": preds})
KH.save_test_predictions(utils.convert_predictions(predictions), estimator, "predictions")
KH.end_pipeline()
return estimator
def run_knn_pipeline():
###### DATA LOADING
xy = loader.XY4(KH) # CAN CHANGE
X = xy["X_train"]
y = xy["y_train"]
X_val = xy["X_val"]
y_val = xy["y_val"]
X_test = xy["X_test"]
output_index = xy["X_test_index"]
print("LOADED DATA")
###### PIPELINE/CV VARIABLES
###### DO NOT CHANGE BEFORE
clf = KNeighborsClassifier()
fl = X.shape[1] # use for n_components
cv_grid = {"clf__metric": ["euclidean", "manhattan"], "clf__n_neighbors": [10, 100, 1000]}
num_folds = 3
####### START PREDICTIONS
print("TRAINING ESTIMATOR")
pred_pipe = Pipeline(steps=[("clf", clf)])
###### DO NOT CHANGE AFTER
estimator = GridSearchCV(pred_pipe, cv_grid, cv=num_folds)
# DO NOT NEED TO CHANGE BEYOND THIS LINE
KH.record_metric("validation", "start", estimator, "training", "", "")
estimator.fit(X, y)
KH.record_metric("validation", "end", estimator, "training", "", "")
KH.record_metric("validation", "end", estimator, "best_params", str(estimator.best_params_), "NA")
KH.record_metric("validation", "end", estimator, "best_estimator", str(estimator.best_estimator_), "NA")
KH.record_metric("validation", "end", estimator, "best_score", str(estimator.best_score_), "NA")
validation_score = str(estimator.score(X_val, y_val))
KH.record_metric("validation", "end", estimator, "validation score", validation_score, "")
preds = estimator.predict(X_test)
predictions = pd.DataFrame({"VisitNumber": output_index, "TripType": preds})
KH.save_test_predictions(utils.convert_predictions(predictions), estimator, "predictions")
KH.end_pipeline()
return estimator
