import copy
import os
import warnings
from collections import OrderedDict
import numpy as np
import pandas as pd
import woodwork as ww
from sklearn.exceptions import NotFittedError
from sklearn.inspection import partial_dependence as sk_partial_dependence
from sklearn.manifold import TSNE
from sklearn.metrics import auc as sklearn_auc
from sklearn.metrics import confusion_matrix as sklearn_confusion_matrix
from sklearn.metrics import \
precision_recall_curve as sklearn_precision_recall_curve
from sklearn.metrics import roc_curve as sklearn_roc_curve
from sklearn.preprocessing import LabelBinarizer
from sklearn.tree import export_graphviz
from sklearn.utils.multiclass import unique_labels
import evalml
from evalml.exceptions import NoPositiveLabelException, NullsInColumnWarning
from evalml.model_family import ModelFamily
from evalml.model_understanding.permutation_importance import (
calculate_permutation_importance
)
from evalml.objectives.utils import get_objective
from evalml.problem_types import ProblemTypes
from evalml.utils import import_or_raise, infer_feature_types, jupyter_check
[docs]def confusion_matrix(y_true, y_predicted, normalize_method='true'):
"""Confusion matrix for binary and multiclass classification.
Arguments:
y_true (pd.Series or np.ndarray): True binary labels.
y_pred (pd.Series or np.ndarray): Predictions from a binary classifier.
normalize_method ({'true', 'pred', 'all', None}): Normalization method to use, if not None. Supported options are: 'true' to normalize by row, 'pred' to normalize by column, or 'all' to normalize by all values. Defaults to 'true'.
Returns:
pd.DataFrame: Confusion matrix. The column header represents the predicted labels while row header represents the actual labels.
"""
y_true = infer_feature_types(y_true)
y_predicted = infer_feature_types(y_predicted)
y_true = y_true.to_numpy()
y_predicted = y_predicted.to_numpy()
labels = unique_labels(y_true, y_predicted)
conf_mat = sklearn_confusion_matrix(y_true, y_predicted)
conf_mat = pd.DataFrame(conf_mat, index=labels, columns=labels)
if normalize_method is not None:
return normalize_confusion_matrix(conf_mat, normalize_method=normalize_method)
return conf_mat
[docs]def normalize_confusion_matrix(conf_mat, normalize_method='true'):
"""Normalizes a confusion matrix.
Arguments:
conf_mat (pd.DataFrame or np.ndarray): Confusion matrix to normalize.
normalize_method ({'true', 'pred', 'all'}): Normalization method. Supported options are: 'true' to normalize by row, 'pred' to normalize by column, or 'all' to normalize by all values. Defaults to 'true'.
Returns:
pd.DataFrame: normalized version of the input confusion matrix. The column header represents the predicted labels while row header represents the actual labels.
"""
conf_mat = infer_feature_types(conf_mat)
col_names = conf_mat.columns
conf_mat = conf_mat.to_numpy()
with warnings.catch_warnings(record=True) as w:
if normalize_method == 'true':
conf_mat = conf_mat.astype('float') / conf_mat.sum(axis=1)[:, np.newaxis]
elif normalize_method == 'pred':
conf_mat = conf_mat.astype('float') / conf_mat.sum(axis=0)
elif normalize_method == 'all':
conf_mat = conf_mat.astype('float') / conf_mat.sum().sum()
else:
raise ValueError('Invalid value provided for "normalize_method": {}'.format(normalize_method))
if w and "invalid value encountered in" in str(w[0].message):
raise ValueError("Sum of given axis is 0 and normalization is not possible. Please select another option.")
conf_mat = pd.DataFrame(conf_mat, index=col_names, columns=col_names)
return conf_mat
[docs]def graph_confusion_matrix(y_true, y_pred, normalize_method='true', title_addition=None):
"""Generate and display a confusion matrix plot.
If `normalize_method` is set, hover text will show raw count, otherwise hover text will show count normalized with method 'true'.
Arguments:
y_true (pd.Series or np.ndarray): True binary labels.
y_pred (pd.Series or np.ndarray): Predictions from a binary classifier.
normalize_method ({'true', 'pred', 'all', None}): Normalization method to use, if not None. Supported options are: 'true' to normalize by row, 'pred' to normalize by column, or 'all' to normalize by all values. Defaults to 'true'.
title_addition (str or None): if not None, append to plot title. Defaults to None.
Returns:
plotly.Figure representing the confusion matrix plot generated
"""
_go = import_or_raise("plotly.graph_objects", error_msg="Cannot find dependency plotly.graph_objects")
_ff = import_or_raise("plotly.figure_factory", error_msg="Cannot find dependency plotly.figure_factory")
if jupyter_check():
import_or_raise("ipywidgets", warning=True)
conf_mat = confusion_matrix(y_true, y_pred, normalize_method=None)
conf_mat_normalized = confusion_matrix(y_true, y_pred, normalize_method=normalize_method or 'true')
labels = conf_mat.columns.tolist()
title = 'Confusion matrix{}{}'.format(
'' if title_addition is None else (' ' + title_addition),
'' if normalize_method is None else (', normalized using method "' + normalize_method + '"'))
z_data, custom_data = (conf_mat, conf_mat_normalized) if normalize_method is None else (conf_mat_normalized, conf_mat)
z_data = z_data.to_numpy()
z_text = [["{:.3f}".format(y) for y in x] for x in z_data]
primary_heading, secondary_heading = ('Raw', 'Normalized') if normalize_method is None else ('Normalized', 'Raw')
hover_text = '<br><b>' + primary_heading + ' Count</b>: %{z}<br><b>' + secondary_heading + ' Count</b>: %{customdata} <br>'
# the "<extra> tags at the end are necessary to remove unwanted trace info
hover_template = '<b>True</b>: %{y}<br><b>Predicted</b>: %{x}' + hover_text + '<extra></extra>'
layout = _go.Layout(title={'text': title},
xaxis={'title': 'Predicted Label', 'type': 'category', 'tickvals': labels},
yaxis={'title': 'True Label', 'type': 'category', 'tickvals': labels})
fig = _ff.create_annotated_heatmap(z_data, x=labels, y=labels,
annotation_text=z_text,
customdata=custom_data,
hovertemplate=hover_template,
colorscale='Blues',
showscale=True)
fig.update_layout(layout)
# put xaxis text on bottom to not overlap with title
fig['layout']['xaxis'].update(side='bottom')
# plotly Heatmap y axis defaults to the reverse of what we want: https://community.plotly.com/t/heatmap-y-axis-is-reversed-by-default-going-against-standard-convention-for-matrices/32180
fig.update_yaxes(autorange="reversed")
return fig
[docs]def precision_recall_curve(y_true, y_pred_proba, pos_label_idx=-1):
"""
Given labels and binary classifier predicted probabilities, compute and return the data representing a precision-recall curve.
Arguments:
y_true (pd.Series or np.ndarray): True binary labels.
y_pred_proba (pd.Series or np.ndarray): Predictions from a binary classifier, before thresholding has been applied. Note this should be the predicted probability for the "true" label.
pos_label_idx (int): the column index corresponding to the positive class. If predicted probabilities are two-dimensional, this will be used to access the probabilities for the positive class.
Returns:
list: Dictionary containing metrics used to generate a precision-recall plot, with the following keys:
* `precision`: Precision values.
* `recall`: Recall values.
* `thresholds`: Threshold values used to produce the precision and recall.
* `auc_score`: The area under the ROC curve.
"""
y_true = infer_feature_types(y_true)
y_pred_proba = infer_feature_types(y_pred_proba)
if isinstance(y_pred_proba, pd.DataFrame):
y_pred_proba_shape = y_pred_proba.shape
try:
y_pred_proba = y_pred_proba.iloc[:, pos_label_idx]
except IndexError:
raise NoPositiveLabelException(f"Predicted probabilities of shape {y_pred_proba_shape} don't contain a column at index {pos_label_idx}")
precision, recall, thresholds = sklearn_precision_recall_curve(y_true, y_pred_proba)
auc_score = sklearn_auc(recall, precision)
return {'precision': precision,
'recall': recall,
'thresholds': thresholds,
'auc_score': auc_score}
[docs]def graph_precision_recall_curve(y_true, y_pred_proba, title_addition=None):
"""Generate and display a precision-recall plot.
Arguments:
y_true (pd.Series or np.ndarray): True binary labels.
y_pred_proba (pd.Series or np.ndarray): Predictions from a binary classifier, before thresholding has been applied. Note this should be the predicted probability for the "true" label.
title_addition (str or None): If not None, append to plot title. Default None.
Returns:
plotly.Figure representing the precision-recall plot generated
"""
_go = import_or_raise("plotly.graph_objects", error_msg="Cannot find dependency plotly.graph_objects")
if jupyter_check():
import_or_raise("ipywidgets", warning=True)
precision_recall_curve_data = precision_recall_curve(y_true, y_pred_proba)
title = 'Precision-Recall{}'.format('' if title_addition is None else (' ' + title_addition))
layout = _go.Layout(title={'text': title},
xaxis={'title': 'Recall', 'range': [-0.05, 1.05]},
yaxis={'title': 'Precision', 'range': [-0.05, 1.05]})
data = []
data.append(_go.Scatter(x=precision_recall_curve_data['recall'], y=precision_recall_curve_data['precision'],
name='Precision-Recall (AUC {:06f})'.format(precision_recall_curve_data['auc_score']),
line=dict(width=3)))
return _go.Figure(layout=layout, data=data)
[docs]def roc_curve(y_true, y_pred_proba):
"""
Given labels and classifier predicted probabilities, compute and return the data representing a Receiver Operating Characteristic (ROC) curve. Works with binary or multiclass problems.
Arguments:
y_true (pd.Series or np.ndarray): True labels.
y_pred_proba (pd.Series or np.ndarray): Predictions from a classifier, before thresholding has been applied.
Returns:
list(dict): A list of dictionaries (with one for each class) is returned. Binary classification problems return a list with one dictionary.
Each dictionary contains metrics used to generate an ROC plot with the following keys:
* `fpr_rate`: False positive rate.
* `tpr_rate`: True positive rate.
* `threshold`: Threshold values used to produce each pair of true/false positive rates.
* `auc_score`: The area under the ROC curve.
"""
y_true = infer_feature_types(y_true).to_numpy()
y_pred_proba = infer_feature_types(y_pred_proba).to_numpy()
if len(y_pred_proba.shape) == 1:
y_pred_proba = y_pred_proba.reshape(-1, 1)
if y_pred_proba.shape[1] == 2:
y_pred_proba = y_pred_proba[:, 1].reshape(-1, 1)
nan_indices = np.logical_or(pd.isna(y_true), np.isnan(y_pred_proba).any(axis=1))
y_true = y_true[~nan_indices]
y_pred_proba = y_pred_proba[~nan_indices]
lb = LabelBinarizer()
lb.fit(np.unique(y_true))
y_one_hot_true = lb.transform(y_true)
n_classes = y_one_hot_true.shape[1]
curve_data = []
for i in range(n_classes):
fpr_rates, tpr_rates, thresholds = sklearn_roc_curve(y_one_hot_true[:, i], y_pred_proba[:, i])
auc_score = sklearn_auc(fpr_rates, tpr_rates)
curve_data.append({'fpr_rates': fpr_rates,
'tpr_rates': tpr_rates,
'thresholds': thresholds,
'auc_score': auc_score})
return curve_data
[docs]def graph_roc_curve(y_true, y_pred_proba, custom_class_names=None, title_addition=None):
"""Generate and display a Receiver Operating Characteristic (ROC) plot for binary and multiclass classification problems.
Arguments:
y_true (pd.Series or np.ndarray): True labels.
y_pred_proba (pd.Series or np.ndarray): Predictions from a classifier, before thresholding has been applied. Note this should a one dimensional array with the predicted probability for the "true" label in the binary case.
custom_class_labels (list or None): If not None, custom labels for classes. Default None.
title_addition (str or None): if not None, append to plot title. Default None.
Returns:
plotly.Figure representing the ROC plot generated
"""
_go = import_or_raise("plotly.graph_objects", error_msg="Cannot find dependency plotly.graph_objects")
if jupyter_check():
import_or_raise("ipywidgets", warning=True)
title = 'Receiver Operating Characteristic{}'.format('' if title_addition is None else (' ' + title_addition))
layout = _go.Layout(title={'text': title},
xaxis={'title': 'False Positive Rate', 'range': [-0.05, 1.05]},
yaxis={'title': 'True Positive Rate', 'range': [-0.05, 1.05]})
all_curve_data = roc_curve(y_true, y_pred_proba)
graph_data = []
n_classes = len(all_curve_data)
if custom_class_names and len(custom_class_names) != n_classes:
raise ValueError('Number of custom class names does not match number of classes')
for i in range(n_classes):
roc_curve_data = all_curve_data[i]
name = i + 1 if custom_class_names is None else custom_class_names[i]
graph_data.append(_go.Scatter(x=roc_curve_data['fpr_rates'], y=roc_curve_data['tpr_rates'],
hovertemplate="(False Postive Rate: %{x}, True Positive Rate: %{y})<br>" + "Threshold: %{text}",
name=f"Class {name} (AUC {roc_curve_data['auc_score']:.06f})",
text=roc_curve_data["thresholds"],
line=dict(width=3)))
graph_data.append(_go.Scatter(x=[0, 1], y=[0, 1],
name='Trivial Model (AUC 0.5)',
line=dict(dash='dash')))
return _go.Figure(layout=layout, data=graph_data)
[docs]def graph_permutation_importance(pipeline, X, y, objective, importance_threshold=0):
"""Generate a bar graph of the pipeline's permutation importance.
Arguments:
pipeline (PipelineBase or subclass): Fitted pipeline
X (pd.DataFrame): The input data used to score and compute permutation importance
y (pd.Series): The target data
objective (str, ObjectiveBase): Objective to score on
importance_threshold (float, optional): If provided, graph features with a permutation importance whose absolute value is larger than importance_threshold. Defaults to zero.
Returns:
plotly.Figure, a bar graph showing features and their respective permutation importance.
"""
go = import_or_raise("plotly.graph_objects", error_msg="Cannot find dependency plotly.graph_objects")
if jupyter_check():
import_or_raise("ipywidgets", warning=True)
perm_importance = calculate_permutation_importance(pipeline, X, y, objective)
perm_importance['importance'] = perm_importance['importance']
if importance_threshold < 0:
raise ValueError(f'Provided importance threshold of {importance_threshold} must be greater than or equal to 0')
# Remove features with close to zero importance
perm_importance = perm_importance[abs(perm_importance['importance']) >= importance_threshold]
# List is reversed to go from ascending order to descending order
perm_importance = perm_importance.iloc[::-1]
title = "Permutation Importance"
subtitle = "The relative importance of each input feature's "\
"overall influence on the pipelines' predictions, computed using "\
"the permutation importance algorithm."
data = [go.Bar(x=perm_importance['importance'],
y=perm_importance['feature'],
orientation='h'
)]
layout = {
'title': '{0}<br><sub>{1}</sub>'.format(title, subtitle),
'height': 800,
'xaxis_title': 'Permutation Importance',
'yaxis_title': 'Feature',
'yaxis': {
'type': 'category'
}
}
fig = go.Figure(data=data, layout=layout)
return fig
[docs]def binary_objective_vs_threshold(pipeline, X, y, objective, steps=100):
"""Computes objective score as a function of potential binary classification
decision thresholds for a fitted binary classification pipeline.
Arguments:
pipeline (BinaryClassificationPipeline obj): Fitted binary classification pipeline
X (pd.DataFrame): The input data used to compute objective score
y (pd.Series): The target labels
objective (ObjectiveBase obj, str): Objective used to score
steps (int): Number of intervals to divide and calculate objective score at
Returns:
pd.DataFrame: DataFrame with thresholds and the corresponding objective score calculated at each threshold
"""
objective = get_objective(objective, return_instance=True)
if not objective.is_defined_for_problem_type(ProblemTypes.BINARY):
raise ValueError("`binary_objective_vs_threshold` can only be calculated for binary classification objectives")
if objective.score_needs_proba:
raise ValueError("Objective `score_needs_proba` must be False")
pipeline_tmp = copy.copy(pipeline)
thresholds = np.linspace(0, 1, steps + 1)
costs = []
for threshold in thresholds:
pipeline_tmp.threshold = threshold
scores = pipeline_tmp.score(X, y, [objective])
costs.append(scores[objective.name])
df = pd.DataFrame({"threshold": thresholds, "score": costs})
return df
[docs]def graph_binary_objective_vs_threshold(pipeline, X, y, objective, steps=100):
"""Generates a plot graphing objective score vs. decision thresholds for a fitted binary classification pipeline.
Arguments:
pipeline (PipelineBase or subclass): Fitted pipeline
X (pd.DataFrame): The input data used to score and compute scores
y (pd.Series): The target labels
objective (ObjectiveBase obj, str): Objective used to score, shown on the y-axis of the graph
steps (int): Number of intervals to divide and calculate objective score at
Returns:
plotly.Figure representing the objective score vs. threshold graph generated
"""
_go = import_or_raise("plotly.graph_objects", error_msg="Cannot find dependency plotly.graph_objects")
if jupyter_check():
import_or_raise("ipywidgets", warning=True)
objective = get_objective(objective, return_instance=True)
df = binary_objective_vs_threshold(pipeline, X, y, objective, steps)
title = f'{objective.name} Scores vs. Thresholds'
layout = _go.Layout(title={'text': title},
xaxis={'title': 'Threshold', 'range': _calculate_axis_range(df['threshold'])},
yaxis={'title': f"{objective.name} Scores vs. Binary Classification Decision Threshold", 'range': _calculate_axis_range(df['score'])})
data = []
data.append(_go.Scatter(x=df['threshold'],
y=df['score'],
line=dict(width=3)))
return _go.Figure(layout=layout, data=data)
def _is_feature_of_type(feature, X, ltype):
"""Determine whether the feature the user passed in to partial dependence is a Woodwork logical type."""
if isinstance(feature, int):
is_type = X.ww.logical_types[X.columns[feature]] == ltype
else:
is_type = X.ww.logical_types[feature] == ltype
return is_type
def _put_categorical_feature_first(features, first_feature_categorical):
"""If the user is doing a two-way partial dependence plot and one of the features is categorical,
we need to make sure the categorical feature is the first element in the tuple that's passed to sklearn.
This is because in the two-way grid calculation, sklearn will try to coerce every element of the grid to the
type of the first feature in the tuple. If we put the categorical feature first, the grid will be of type 'object'
which can accommodate both categorical and numeric data. If we put the numeric feature first, the grid will be of
type float64 and we can't coerce categoricals to float64 dtype.
"""
new_features = features if first_feature_categorical else (features[1], features[0])
return new_features
def _get_feature_names_from_str_or_col_index(X, names_or_col_indices):
"""Helper function to map the user-input features param to column names."""
feature_list = []
for name_or_index in names_or_col_indices:
if isinstance(name_or_index, int):
feature_list.append(X.columns[name_or_index])
else:
feature_list.append(name_or_index)
return feature_list
def _raise_value_error_if_any_features_all_nan(df):
"""Helper for partial dependence data validation."""
nan_pct = df.isna().mean()
all_nan = nan_pct[nan_pct == 1].index.tolist()
all_nan = [f"'{name}'" for name in all_nan]
if all_nan:
raise ValueError("The following features have all NaN values and so the "
f"partial dependence cannot be computed: {', '.join(all_nan)}")
def _raise_value_error_if_mostly_one_value(df, percentile):
"""Helper for partial dependence data validation."""
one_value = []
values = []
for col in df.columns:
normalized_counts = df[col].value_counts(normalize=True) + 0.01
normalized_counts = normalized_counts[normalized_counts > percentile]
if not normalized_counts.empty:
one_value.append(f"'{col}'")
values.append(str(normalized_counts.index[0]))
if one_value:
raise ValueError(f"Features ({', '.join(one_value)}) are mostly one value, ({', '.join(values)}), "
f"and cannot be used to compute partial dependence. Try raising the upper percentage value.")
[docs]def partial_dependence(pipeline, X, features, percentiles=(0.05, 0.95), grid_resolution=100):
"""Calculates one or two-way partial dependence. If a single integer or
string is given for features, one-way partial dependence is calculated. If
a tuple of two integers or strings is given, two-way partial dependence
is calculated with the first feature in the y-axis and second feature in the
x-axis.
Arguments:
pipeline (PipelineBase or subclass): Fitted pipeline
X (pd.DataFrame, np.ndarray): The input data used to generate a grid of values
for feature where partial dependence will be calculated at
features (int, string, tuple[int or string]): The target feature for which to create the partial dependence plot for.
If features is an int, it must be the index of the feature to use.
If features is a string, it must be a valid column name in X.
If features is a tuple of int/strings, it must contain valid column integers/names in X.
percentiles (tuple[float]): The lower and upper percentile used to create the extreme values for the grid.
Must be in [0, 1]. Defaults to (0.05, 0.95).
grid_resolution (int): Number of samples of feature(s) for partial dependence plot. If this value
is less than the maximum number of categories present in categorical data within X, it will be
set to the max number of categories + 1. Defaults to 100.
Returns:
pd.DataFrame: DataFrame with averaged predictions for all points in the grid averaged
over all samples of X and the values used to calculate those predictions.
In the one-way case: The dataframe will contain two columns, "feature_values" (grid points at which the
partial dependence was calculated) and "partial_dependence" (the partial dependence at that feature value).
For classification problems, there will be a third column called "class_label" (the class label for which
the partial dependence was calculated). For binary classification, the partial dependence is only calculated
for the "positive" class.
In the two-way case: The data frame will contain grid_resolution number of columns and rows where the
index and column headers are the sampled values of the first and second features, respectively, used to make
the partial dependence contour. The values of the data frame contain the partial dependence data for each
feature value pair.
Raises:
ValueError: if the user provides a tuple of not exactly two features.
ValueError: if the provided pipeline isn't fitted.
ValueError: if the provided pipeline is a Baseline pipeline.
ValueError: if any of the features passed in are completely NaN
ValueError: if any of the features are low-variance. Defined as having one value occurring more than the upper
percentile passed by the user. By default 95%.
"""
# Dynamically set the grid resolution to the maximum number of values
# in the categorical/datetime variables if there are more categories/datetime values than resolution cells
X = infer_feature_types(X)
if isinstance(features, (list, tuple)):
is_categorical = [_is_feature_of_type(f, X, ww.logical_types.Categorical) for f in features]
is_datetime = [_is_feature_of_type(f, X, ww.logical_types.Datetime) for f in features]
else:
is_categorical = [_is_feature_of_type(features, X, ww.logical_types.Categorical)]
is_datetime = [_is_feature_of_type(features, X, ww.logical_types.Datetime)]
if isinstance(features, (list, tuple)):
if len(features) != 2:
raise ValueError("Too many features given to graph_partial_dependence. Only one or two-way partial "
"dependence is supported.")
if not (all([isinstance(x, str) for x in features]) or all([isinstance(x, int) for x in features])):
raise ValueError("Features provided must be a tuple entirely of integers or strings, not a mixture of both.")
X_features = X.ww.iloc[:, list(features)] if isinstance(features[0], int) else X.ww[list(features)]
else:
X_features = X.ww.iloc[:, [features]] if isinstance(features, int) else X.ww[[features]]
X_cats = X_features.ww.select("categorical")
if any(is_categorical):
max_num_cats = max(X_cats.ww.describe().loc["nunique"])
grid_resolution = max([max_num_cats + 1, grid_resolution])
X_dt = X_features.ww.select("datetime")
if any(is_datetime):
max_num_dt = max(X_dt.ww.describe().loc["nunique"])
grid_resolution = max([max_num_dt + 1, grid_resolution])
if isinstance(features, (list, tuple)):
feature_names = _get_feature_names_from_str_or_col_index(X, features)
if any(is_datetime):
raise ValueError('Two-way partial dependence is not supported for datetime columns.')
if any(is_categorical):
features = _put_categorical_feature_first(features, is_categorical[0])
else:
feature_names = _get_feature_names_from_str_or_col_index(X, [features])
if not pipeline._is_fitted:
raise ValueError("Pipeline to calculate partial dependence for must be fitted")
if pipeline.model_family == ModelFamily.BASELINE:
raise ValueError("Partial dependence plots are not supported for Baseline pipelines")
feature_list = X[feature_names]
_raise_value_error_if_any_features_all_nan(feature_list)
if feature_list.isnull().sum().any():
warnings.warn("There are null values in the features, which will cause NaN values in the partial dependence output. "
"Fill in these values to remove the NaN values.", NullsInColumnWarning)
_raise_value_error_if_mostly_one_value(feature_list, percentiles[1])
wrapped = evalml.pipelines.components.utils.scikit_learn_wrapped_estimator(pipeline)
avg_pred, values = sk_partial_dependence(wrapped, X=X, features=features, percentiles=percentiles, grid_resolution=grid_resolution)
classes = None
if isinstance(pipeline, evalml.pipelines.BinaryClassificationPipeline):
classes = [pipeline.classes_[1]]
elif isinstance(pipeline, evalml.pipelines.MulticlassClassificationPipeline):
classes = pipeline.classes_
if isinstance(features, (int, str)):
data = pd.DataFrame({"feature_values": np.tile(values[0], avg_pred.shape[0]),
"partial_dependence": np.concatenate([pred for pred in avg_pred])})
elif isinstance(features, (list, tuple)):
data = pd.DataFrame(avg_pred.reshape((-1, avg_pred.shape[-1])))
data.columns = values[1]
data.index = np.tile(values[0], avg_pred.shape[0])
if classes is not None:
data['class_label'] = np.repeat(classes, len(values[0]))
return data
def _update_fig_with_two_way_partial_dependence(_go, fig, label_df, part_dep, features, is_categorical,
label=None, row=None, col=None):
"""Helper for formatting the two-way partial dependence plot."""
y = label_df.index
x = label_df.columns
z = label_df.values
if not any(is_categorical):
# No features are categorical. In this case, we pass both x and y data to the Contour plot so that
# plotly can figure out the axis formatting for us.
kwargs = {"x": x, "y": y}
fig.update_xaxes(title=f'{features[1]}',
range=_calculate_axis_range(np.array([x for x in part_dep.columns if x != 'class_label'])),
row=row, col=col)
fig.update_yaxes(range=_calculate_axis_range(part_dep.index), row=row, col=col)
elif sum(is_categorical) == 1:
# One feature is categorical. Since we put the categorical feature first, the numeric feature will be the x
# axis. So we pass the x to the Contour plot so that plotly can format it for us.
# Since the y axis is a categorical value, we will set the y tickmarks ourselves. Passing y to the contour plot
# in this case will "work" but the formatting will look bad.
kwargs = {"x": x}
fig.update_xaxes(title=f'{features[1]}',
range=_calculate_axis_range(np.array([x for x in part_dep.columns if x != 'class_label'])),
row=row, col=col)
fig.update_yaxes(tickmode='array', tickvals=list(range(label_df.shape[0])),
ticktext=list(label_df.index), row=row, col=col)
else:
# Both features are categorical so we must format both axes ourselves.
kwargs = {}
fig.update_yaxes(tickmode='array', tickvals=list(range(label_df.shape[0])),
ticktext=list(label_df.index), row=row, col=col)
fig.update_xaxes(tickmode='array', tickvals=list(range(label_df.shape[1])),
ticktext=list(label_df.columns), row=row, col=col)
fig.add_trace(_go.Contour(z=z, name=label, coloraxis="coloraxis", **kwargs), row=row, col=col)
[docs]def graph_partial_dependence(pipeline, X, features, class_label=None, grid_resolution=100):
"""Create an one-way or two-way partial dependence plot. Passing a single integer or
string as features will create a one-way partial dependence plot with the feature values
plotted against the partial dependence. Passing features a tuple of int/strings will create
a two-way partial dependence plot with a contour of feature[0] in the y-axis, feature[1]
in the x-axis and the partial dependence in the z-axis.
Arguments:
pipeline (PipelineBase or subclass): Fitted pipeline
X (pd.DataFrame, np.ndarray): The input data used to generate a grid of values
for feature where partial dependence will be calculated at
features (int, string, tuple[int or string]): The target feature for which to create the partial dependence plot for.
If features is an int, it must be the index of the feature to use.
If features is a string, it must be a valid column name in X.
If features is a tuple of strings, it must contain valid column int/names in X.
class_label (string, optional): Name of class to plot for multiclass problems. If None, will plot
the partial dependence for each class. This argument does not change behavior for regression or binary
classification pipelines. For binary classification, the partial dependence for the positive label will
always be displayed. Defaults to None.
grid_resolution (int): Number of samples of feature(s) for partial dependence plot
Returns:
plotly.graph_objects.Figure: figure object containing the partial dependence data for plotting
Raises:
ValueError: if a graph is requested for a class name that isn't present in the pipeline
"""
X = infer_feature_types(X)
if isinstance(features, (list, tuple)):
mode = "two-way"
is_categorical = [_is_feature_of_type(f, X, ww.logical_types.Categorical) for f in features]
if any(is_categorical):
features = _put_categorical_feature_first(features, is_categorical[0])
elif isinstance(features, (int, str)):
mode = "one-way"
is_categorical = _is_feature_of_type(features, X, ww.logical_types.Categorical)
_go = import_or_raise("plotly.graph_objects", error_msg="Cannot find dependency plotly.graph_objects")
if jupyter_check():
import_or_raise("ipywidgets", warning=True)
if isinstance(pipeline, evalml.pipelines.MulticlassClassificationPipeline) and class_label is not None:
if class_label not in pipeline.classes_:
msg = f"Class {class_label} is not one of the classes the pipeline was fit on: {', '.join(list(pipeline.classes_))}"
raise ValueError(msg)
part_dep = partial_dependence(pipeline, X, features=features, grid_resolution=grid_resolution)
if mode == "two-way":
title = f"Partial Dependence of '{features[0]}' vs. '{features[1]}'"
layout = _go.Layout(title={'text': title},
xaxis={'title': f'{features[1]}'},
yaxis={'title': f'{features[0]}'},
showlegend=False)
elif mode == "one-way":
feature_name = str(features)
title = f"Partial Dependence of '{feature_name}'"
layout = _go.Layout(title={'text': title},
xaxis={'title': f'{feature_name}'},
yaxis={'title': 'Partial Dependence'},
showlegend=False)
if isinstance(pipeline, evalml.pipelines.MulticlassClassificationPipeline):
class_labels = [class_label] if class_label is not None else pipeline.classes_
_subplots = import_or_raise("plotly.subplots", error_msg="Cannot find dependency plotly.graph_objects")
# If the user passes in a value for class_label, we want to create a 1 x 1 subplot or else there would
# be an empty column in the plot and it would look awkward
rows, cols = ((len(class_labels) + 1) // 2, 2) if len(class_labels) > 1 else (1, len(class_labels))
# Don't specify share_xaxis and share_yaxis so that we get tickmarks in each subplot
fig = _subplots.make_subplots(rows=rows, cols=cols, subplot_titles=class_labels)
for i, label in enumerate(class_labels):
label_df = part_dep.loc[part_dep.class_label == label]
row = (i + 2) // 2
col = (i % 2) + 1
label_df.drop(columns=['class_label'], inplace=True)
if mode == 'two-way':
_update_fig_with_two_way_partial_dependence(_go, fig, label_df, part_dep, features, is_categorical,
label, row, col)
elif mode == "one-way":
x = label_df['feature_values']
y = label_df['partial_dependence']
if is_categorical:
trace = _go.Bar(x=x, y=y, name=label)
else:
trace = _go.Scatter(x=x, y=y, line=dict(width=3), name=label)
fig.add_trace(trace, row=row, col=col)
fig.update_layout(layout)
if mode == "two-way":
fig.update_layout(coloraxis=dict(colorscale='Bluered_r'), showlegend=False)
elif mode == "one-way":
title = f'{feature_name}'
xrange = _calculate_axis_range(part_dep['feature_values']) if not is_categorical else None
yrange = _calculate_axis_range(part_dep['partial_dependence'])
fig.update_xaxes(title=title, range=xrange)
fig.update_yaxes(range=yrange)
return fig
else:
if "class_label" in part_dep.columns:
part_dep.drop(columns=['class_label'], inplace=True)
if mode == "two-way":
fig = _go.Figure(layout=layout)
_update_fig_with_two_way_partial_dependence(_go, fig, part_dep, part_dep, features, is_categorical,
label="Partial Dependence", row=None, col=None)
return fig
elif mode == "one-way":
if is_categorical:
trace = _go.Bar(x=part_dep['feature_values'], y=part_dep['partial_dependence'],
name="Partial Dependence")
else:
trace = _go.Scatter(x=part_dep['feature_values'],
y=part_dep['partial_dependence'],
name='Partial Dependence',
line=dict(width=3))
return _go.Figure(layout=layout, data=[trace])
def _calculate_axis_range(arr):
"""Helper method to help calculate the appropriate range for an axis based on the data to graph."""
max_value = arr.max()
min_value = arr.min()
margins = abs(max_value - min_value) * 0.05
return [min_value - margins, max_value + margins]
[docs]def get_prediction_vs_actual_data(y_true, y_pred, outlier_threshold=None):
"""Combines y_true and y_pred into a single dataframe and adds a column for outliers. Used in `graph_prediction_vs_actual()`.
Arguments:
y_true (pd.Series, or np.ndarray): The real target values of the data
y_pred (pd.Series, or np.ndarray): The predicted values outputted by the regression model.
outlier_threshold (int, float): A positive threshold for what is considered an outlier value. This value is compared to the absolute difference
between each value of y_true and y_pred. Values within this threshold will be blue, otherwise they will be yellow.
Defaults to None
Returns:
pd.DataFrame with the following columns:
* `prediction`: Predicted values from regression model.
* `actual`: Real target values.
* `outlier`: Colors indicating which values are in the threshold for what is considered an outlier value.
"""
if outlier_threshold and outlier_threshold <= 0:
raise ValueError(f"Threshold must be positive! Provided threshold is {outlier_threshold}")
y_true = infer_feature_types(y_true)
y_pred = infer_feature_types(y_pred)
predictions = y_pred.reset_index(drop=True)
actual = y_true.reset_index(drop=True)
data = pd.concat([pd.Series(predictions),
pd.Series(actual)], axis=1)
data.columns = ['prediction', 'actual']
if outlier_threshold:
data['outlier'] = np.where((abs(data['prediction'] - data['actual']) >= outlier_threshold), "#ffff00", "#0000ff")
else:
data['outlier'] = '#0000ff'
return data
[docs]def graph_prediction_vs_actual(y_true, y_pred, outlier_threshold=None):
"""Generate a scatter plot comparing the true and predicted values. Used for regression plotting
Arguments:
y_true (pd.Series): The real target values of the data
y_pred (pd.Series): The predicted values outputted by the regression model.
outlier_threshold (int, float): A positive threshold for what is considered an outlier value. This value is compared to the absolute difference
between each value of y_true and y_pred. Values within this threshold will be blue, otherwise they will be yellow.
Defaults to None
Returns:
plotly.Figure representing the predicted vs. actual values graph
"""
_go = import_or_raise("plotly.graph_objects", error_msg="Cannot find dependency plotly.graph_objects")
if jupyter_check():
import_or_raise("ipywidgets", warning=True)
if outlier_threshold and outlier_threshold <= 0:
raise ValueError(f"Threshold must be positive! Provided threshold is {outlier_threshold}")
df = get_prediction_vs_actual_data(y_true, y_pred, outlier_threshold)
data = []
x_axis = _calculate_axis_range(df['prediction'])
y_axis = _calculate_axis_range(df['actual'])
x_y_line = [min(x_axis[0], y_axis[0]), max(x_axis[1], y_axis[1])]
data.append(_go.Scatter(x=x_y_line, y=x_y_line, name="y = x line", line_color='grey'))
title = 'Predicted vs Actual Values Scatter Plot'
layout = _go.Layout(title={'text': title},
xaxis={'title': 'Prediction', 'range': x_y_line},
yaxis={'title': 'Actual', 'range': x_y_line})
for color, outlier_group in df.groupby('outlier'):
if outlier_threshold:
name = "< outlier_threshold" if color == "#0000ff" else ">= outlier_threshold"
else:
name = "Values"
data.append(_go.Scatter(x=outlier_group['prediction'],
y=outlier_group['actual'],
mode='markers',
marker=_go.scatter.Marker(color=color),
name=name))
return _go.Figure(layout=layout, data=data)
def _tree_parse(est, feature_names):
children_left = est.tree_.children_left
children_right = est.tree_.children_right
features = est.tree_.feature
thresholds = est.tree_.threshold
values = est.tree_.value
def recurse(i):
if children_left[i] == children_right[i]:
return {'Value': values[i]}
return OrderedDict({
'Feature': feature_names[features[i]],
'Threshold': thresholds[i],
'Value': values[i],
'Left_Child': recurse(children_left[i]),
'Right_Child': recurse(children_right[i])
})
return recurse(0)
def decision_tree_data_from_estimator(estimator):
"""Return data for a fitted tree in a restructured format
Arguments:
estimator (ComponentBase): A fitted DecisionTree-based estimator.
Returns:
OrderedDict: An OrderedDict of OrderedDicts describing a tree structure
"""
if not estimator.model_family == ModelFamily.DECISION_TREE:
raise ValueError("Tree structure reformatting is only supported for decision tree estimators")
if not estimator._is_fitted:
raise NotFittedError("This DecisionTree estimator is not fitted yet. Call 'fit' with appropriate arguments "
"before using this estimator.")
est = estimator._component_obj
feature_names = estimator.input_feature_names
return _tree_parse(est, feature_names)
def decision_tree_data_from_pipeline(pipeline_):
"""Return data for a fitted pipeline with in a restructured format
Arguments:
pipeline_ (PipelineBase): A pipeline with a DecisionTree-based estimator.
Returns:
OrderedDict: An OrderedDict of OrderedDicts describing a tree structure
"""
if not pipeline_.model_family == ModelFamily.DECISION_TREE:
raise ValueError("Tree structure reformatting is only supported for decision tree estimators")
if not pipeline_._is_fitted:
raise NotFittedError("The DecisionTree estimator associated with this pipeline is not fitted yet. Call 'fit' "
"with appropriate arguments before using this estimator.")
est = pipeline_.estimator._component_obj
feature_names = pipeline_.input_feature_names[pipeline_.estimator.name]
return _tree_parse(est, feature_names)
def visualize_decision_tree(estimator, max_depth=None, rotate=False, filled=False, filepath=None):
"""Generate an image visualizing the decision tree
Arguments:
estimator (ComponentBase): A fitted DecisionTree-based estimator.
max_depth (int, optional): The depth to which the tree should be displayed. If set to None (as by default),
tree is fully generated.
rotate (bool, optional): Orient tree left to right rather than top-down.
filled (bool, optional): Paint nodes to indicate majority class for classification, extremity of values for
regression, or purity of node for multi-output.
filepath (str, optional): Path to where the graph should be saved. If set to None (as by default), the graph
will not be saved.
Returns:
graphviz.Source: DOT object that can be directly displayed in Jupyter notebooks.
"""
if not estimator.model_family == ModelFamily.DECISION_TREE:
raise ValueError("Tree visualizations are only supported for decision tree estimators")
if max_depth and (not isinstance(max_depth, int) or not max_depth >= 0):
raise ValueError("Unknown value: '{}'. The parameter max_depth has to be a non-negative integer"
.format(max_depth))
if not estimator._is_fitted:
raise NotFittedError("This DecisionTree estimator is not fitted yet. Call 'fit' with appropriate arguments before using this estimator.")
est = estimator._component_obj
graphviz = import_or_raise('graphviz', error_msg='Please install graphviz to visualize trees.')
graph_format = None
if filepath:
# Cast to str in case a Path object was passed in
filepath = str(filepath)
try:
f = open(filepath, 'w')
f.close()
except (IOError, FileNotFoundError):
raise ValueError(('Specified filepath is not writeable: {}'.format(filepath)))
path_and_name, graph_format = os.path.splitext(filepath)
if graph_format:
graph_format = graph_format[1:].lower() # ignore the dot
supported_filetypes = graphviz.backend.FORMATS
if graph_format not in supported_filetypes:
raise ValueError(("Unknown format '{}'. Make sure your format is one of the " +
"following: {}").format(graph_format, supported_filetypes))
else:
graph_format = 'pdf' # If the filepath has no extension default to pdf
dot_data = export_graphviz(decision_tree=est, max_depth=max_depth, rotate=rotate, filled=filled, feature_names=estimator.input_feature_names)
source_obj = graphviz.Source(source=dot_data, format=graph_format)
if filepath:
source_obj.render(filename=path_and_name, cleanup=True)
return source_obj
[docs]def get_prediction_vs_actual_over_time_data(pipeline, X, y, dates):
"""Get the data needed for the prediction_vs_actual_over_time plot.
Arguments:
pipeline (TimeSeriesRegressionPipeline): Fitted time series regression pipeline.
X (pd.DataFrame): Features used to generate new predictions.
y (pd.Series): Target values to compare predictions against.
dates (pd.Series): Dates corresponding to target values and predictions.
Returns:
pd.DataFrame
"""
dates = infer_feature_types(dates)
y = infer_feature_types(y)
prediction = pipeline.predict(X, y)
return pd.DataFrame({"dates": dates.reset_index(drop=True),
"target": y.reset_index(drop=True),
"prediction": prediction.reset_index(drop=True)})
[docs]def graph_prediction_vs_actual_over_time(pipeline, X, y, dates):
"""Plot the target values and predictions against time on the x-axis.
Arguments:
pipeline (TimeSeriesRegressionPipeline): Fitted time series regression pipeline.
X (pd.DataFrame): Features used to generate new predictions.
y (pd.Series): Target values to compare predictions against.
dates (pd.Series): Dates corresponding to target values and predictions.
Returns:
plotly.Figure: Showing the prediction vs actual over time.
"""
_go = import_or_raise("plotly.graph_objects", error_msg="Cannot find dependency plotly.graph_objects")
if pipeline.problem_type != ProblemTypes.TIME_SERIES_REGRESSION:
raise ValueError("graph_prediction_vs_actual_over_time only supports time series regression pipelines! "
f"Received {str(pipeline.problem_type)}.")
data = get_prediction_vs_actual_over_time_data(pipeline, X, y, dates)
data = [_go.Scatter(x=data["dates"], y=data["target"], mode='lines+markers', name="Target",
line=dict(color='#1f77b4')),
_go.Scatter(x=data["dates"], y=data["prediction"], mode='lines+markers', name='Prediction',
line=dict(color='#d62728'))]
# Let plotly pick the best date format.
layout = _go.Layout(title={'text': "Prediction vs Target over time"},
xaxis={'title': 'Time'},
yaxis={'title': 'Target Values and Predictions'})
return _go.Figure(data=data, layout=layout)
[docs]def get_linear_coefficients(estimator, features=None):
"""Returns a dataframe showing the features with the greatest predictive power for a linear model.
Arguments:
estimator (Estimator): Fitted linear model family estimator.
features (list[str]): List of feature names associated with the underlying data.
Returns:
pd.DataFrame: Displaying the features by importance.
"""
if not estimator.model_family == ModelFamily.LINEAR_MODEL:
raise ValueError("Linear coefficients are only available for linear family models")
if not estimator._is_fitted:
raise NotFittedError("This linear estimator is not fitted yet. Call 'fit' with appropriate arguments "
"before using this estimator.")
coef_ = estimator.feature_importance
coef_ = pd.Series(coef_, name='Coefficients', index=features)
coef_ = coef_.sort_values()
coef_ = pd.Series(estimator._component_obj.intercept_, index=['Intercept']).append(coef_)
return coef_
[docs]def t_sne(X, n_components=2, perplexity=30.0, learning_rate=200.0, metric='euclidean', **kwargs):
"""Get the transformed output after fitting X to the embedded space using t-SNE.
Arguments:
X (np.ndarray, pd.DataFrame): Data to be transformed. Must be numeric.
n_components (int, optional): Dimension of the embedded space.
perplexity (float, optional): Related to the number of nearest neighbors that is used in other manifold learning
algorithms. Larger datasets usually require a larger perplexity. Consider selecting a value between 5 and 50.
learning_rate (float, optional): Usually in the range [10.0, 1000.0]. If the cost function gets stuck in a bad
local minimum, increasing the learning rate may help.
metric (str, optional): The metric to use when calculating distance between instances in a feature array.
Returns:
np.ndarray (n_samples, n_components)
"""
if not isinstance(n_components, int) or not n_components > 0:
raise ValueError("The parameter n_components must be of type integer and greater than 0")
if not perplexity >= 0:
raise ValueError("The parameter perplexity must be non-negative")
X = infer_feature_types(X)
t_sne_ = TSNE(n_components=n_components, perplexity=perplexity, learning_rate=learning_rate, metric=metric, **kwargs)
X_new = t_sne_.fit_transform(X)
return X_new
[docs]def graph_t_sne(X, n_components=2, perplexity=30.0, learning_rate=200.0, metric='euclidean', marker_line_width=2, marker_size=7, **kwargs):
"""Plot high dimensional data into lower dimensional space using t-SNE .
Arguments:
X (np.ndarray, pd.DataFrame): Data to be transformed. Must be numeric.
n_components (int, optional): Dimension of the embedded space.
perplexity (float, optional): Related to the number of nearest neighbors that is used in other manifold learning
algorithms. Larger datasets usually require a larger perplexity. Consider selecting a value between 5 and 50.
learning_rate (float, optional): Usually in the range [10.0, 1000.0]. If the cost function gets stuck in a bad
local minimum, increasing the learning rate may help.
metric (str, optional): The metric to use when calculating distance between instances in a feature array.
marker_line_width (int, optional): Determines the line width of the marker boundary.
marker_size (int, optional): Determines the size of the marker.
Returns:
plotly.Figure representing the transformed data
"""
_go = import_or_raise("plotly.graph_objects", error_msg="Cannot find dependency plotly.graph_objects")
if not marker_line_width >= 0:
raise ValueError("The parameter marker_line_width must be non-negative")
if not marker_size >= 0:
raise ValueError("The parameter marker_size must be non-negative")
X_embedded = t_sne(X, n_components=n_components, perplexity=perplexity, learning_rate=learning_rate, metric=metric, **kwargs)
fig = _go.Figure()
fig.add_trace(_go.Scatter(
x=X_embedded[:, 0], y=X_embedded[:, 1],
mode='markers'
))
fig.update_traces(mode='markers', marker_line_width=marker_line_width, marker_size=marker_size)
fig.update_layout(title='t-SNE', yaxis_zeroline=False, xaxis_zeroline=False)
return fig