Model Understanding¶
Simply examining a model’s performance metrics is not enough to select a model and promote it for use in a production setting. While developing an ML algorithm, it is important to understand how the model behaves on the data, to examine the key factors influencing its predictions and to consider where it may be deficient. Determination of what “success” may mean for an ML project depends first and foremost on the user’s domain expertise.
EvalML includes a variety of tools for understanding models, from graphing utilities to methods for explaining predictions.
** Graphing methods on Jupyter Notebook and Jupyter Lab require ipywidgets to be installed.
** If graphing on Jupyter Lab, jupyterlab-plotly required. To download this, make sure you have npm installed.
Graphing Utilities¶
First, let’s train a pipeline on some data.
[1]:
import evalml
from evalml.pipelines import BinaryClassificationPipeline
X, y = evalml.demos.load_breast_cancer()
pipeline = BinaryClassificationPipeline(['Simple Imputer', 'Random Forest Classifier'])
pipeline.fit(X, y)
print(pipeline.score(X, y, objectives=['log loss binary']))
OrderedDict([('Log Loss Binary', 0.038403828027876195)])
Feature Importance¶
We can get the importance associated with each feature of the resulting pipeline
[2]:
pipeline.feature_importance
[2]:
| feature | importance | |
|---|---|---|
| 0 | worst perimeter | 0.176488 |
| 1 | worst concave points | 0.125260 |
| 2 | worst radius | 0.124161 |
| 3 | mean concave points | 0.086443 |
| 4 | worst area | 0.072465 |
| 5 | mean concavity | 0.072320 |
| 6 | mean perimeter | 0.056685 |
| 7 | mean area | 0.049599 |
| 8 | area error | 0.037229 |
| 9 | worst concavity | 0.028181 |
| 10 | mean radius | 0.023294 |
| 11 | radius error | 0.019457 |
| 12 | worst texture | 0.014990 |
| 13 | perimeter error | 0.014103 |
| 14 | mean texture | 0.013618 |
| 15 | worst compactness | 0.011310 |
| 16 | worst smoothness | 0.011139 |
| 17 | worst fractal dimension | 0.008118 |
| 18 | worst symmetry | 0.007818 |
| 19 | mean smoothness | 0.006152 |
| 20 | concave points error | 0.005887 |
| 21 | fractal dimension error | 0.005059 |
| 22 | concavity error | 0.004510 |
| 23 | smoothness error | 0.004493 |
| 24 | texture error | 0.004476 |
| 25 | mean compactness | 0.004050 |
| 26 | compactness error | 0.003559 |
| 27 | mean symmetry | 0.003243 |
| 28 | symmetry error | 0.003124 |
| 29 | mean fractal dimension | 0.002768 |
We can also create a bar plot of the feature importances
[3]:
pipeline.graph_feature_importance()
Permutation Importance¶
We can also compute and plot the permutation importance of the pipeline.
[4]:
from evalml.model_understanding import calculate_permutation_importance
calculate_permutation_importance(pipeline, X, y, 'log loss binary')
[4]:
| feature | importance | |
|---|---|---|
| 0 | worst perimeter | 0.083152 |
| 1 | worst radius | 0.078690 |
| 2 | worst area | 0.071237 |
| 3 | worst concave points | 0.071188 |
| 4 | mean concave points | 0.043834 |
| 5 | worst concavity | 0.040660 |
| 6 | mean concavity | 0.039079 |
| 7 | area error | 0.037576 |
| 8 | mean area | 0.027190 |
| 9 | mean perimeter | 0.026886 |
| 10 | worst texture | 0.017269 |
| 11 | mean texture | 0.013273 |
| 12 | perimeter error | 0.011904 |
| 13 | mean radius | 0.011215 |
| 14 | radius error | 0.011004 |
| 15 | worst compactness | 0.009072 |
| 16 | worst smoothness | 0.008203 |
| 17 | mean smoothness | 0.005717 |
| 18 | worst symmetry | 0.004561 |
| 19 | worst fractal dimension | 0.004273 |
| 20 | concavity error | 0.004138 |
| 21 | compactness error | 0.003855 |
| 22 | concave points error | 0.003221 |
| 23 | mean compactness | 0.003207 |
| 24 | smoothness error | 0.002949 |
| 25 | fractal dimension error | 0.002712 |
| 26 | texture error | 0.002541 |
| 27 | mean fractal dimension | 0.002305 |
| 28 | symmetry error | 0.002077 |
| 29 | mean symmetry | 0.001675 |
[5]:
from evalml.model_understanding import graph_permutation_importance
graph_permutation_importance(pipeline, X, y, 'log loss binary')
Partial Dependence Plots¶
We can calculate the one-way partial dependence plots for a feature.
[6]:
from evalml.model_understanding.graphs import partial_dependence
partial_dependence(pipeline, X, features='mean radius')
[6]:
| feature_values | partial_dependence | class_label | |
|---|---|---|---|
| 0 | 9.498540 | 0.371141 | malignant |
| 1 | 9.610488 | 0.371141 | malignant |
| 2 | 9.722436 | 0.371141 | malignant |
| 3 | 9.834384 | 0.371141 | malignant |
| 4 | 9.946332 | 0.371141 | malignant |
| ... | ... | ... | ... |
| 95 | 20.133608 | 0.399560 | malignant |
| 96 | 20.245556 | 0.399560 | malignant |
| 97 | 20.357504 | 0.399560 | malignant |
| 98 | 20.469452 | 0.399560 | malignant |
| 99 | 20.581400 | 0.399560 | malignant |
100 rows × 3 columns
[7]:
from evalml.model_understanding.graphs import graph_partial_dependence
graph_partial_dependence(pipeline, X, features='mean radius')
You can also compute the partial dependence for a categorical feature. We will demonstrate this on the fraud dataset.
[8]:
X_fraud, y_fraud = evalml.demos.load_fraud(100, verbose=False)
X_fraud.ww.init(logical_types={"provider": "Categorical", 'region': "Categorical"})
fraud_pipeline = BinaryClassificationPipeline(["DateTime Featurization Component","One Hot Encoder", "Random Forest Classifier"])
fraud_pipeline.fit(X_fraud, y_fraud)
graph_partial_dependence(fraud_pipeline, X_fraud, features='provider')
Two-way partial dependence plots are also possible and invoke the same API.
[9]:
partial_dependence(pipeline, X, features=('worst perimeter', 'worst radius'), grid_resolution=10)
[9]:
| 10.5072 | 12.193377777777776 | 13.879555555555555 | 15.565733333333334 | 17.251911111111113 | 18.938088888888892 | 20.624266666666667 | 22.310444444444443 | 23.99662222222222 | 25.6828 | class_label | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 67.733600 | 0.264908 | 0.267211 | 0.274328 | 0.286943 | 0.405865 | 0.442701 | 0.444406 | 0.444406 | 0.444406 | 0.444406 | malignant |
| 79.363867 | 0.265840 | 0.268142 | 0.275260 | 0.287875 | 0.405865 | 0.442701 | 0.444406 | 0.444406 | 0.444406 | 0.444406 | malignant |
| 90.994133 | 0.273397 | 0.275699 | 0.282817 | 0.295432 | 0.411805 | 0.448641 | 0.450346 | 0.450346 | 0.450346 | 0.450346 | malignant |
| 102.624400 | 0.298379 | 0.300681 | 0.307799 | 0.323472 | 0.436371 | 0.473207 | 0.474911 | 0.474911 | 0.474911 | 0.474911 | malignant |
| 114.254667 | 0.395976 | 0.398278 | 0.404798 | 0.417739 | 0.530867 | 0.567702 | 0.571516 | 0.571797 | 0.571797 | 0.571797 | malignant |
| 125.884933 | 0.426266 | 0.428569 | 0.435089 | 0.450433 | 0.556594 | 0.593430 | 0.597244 | 0.597525 | 0.597525 | 0.597525 | malignant |
| 137.515200 | 0.442004 | 0.444307 | 0.450827 | 0.466171 | 0.574301 | 0.611137 | 0.614950 | 0.615232 | 0.615232 | 0.615232 | malignant |
| 149.145467 | 0.442004 | 0.444307 | 0.450827 | 0.466171 | 0.574301 | 0.611137 | 0.614950 | 0.615232 | 0.615232 | 0.615232 | malignant |
| 160.775733 | 0.442004 | 0.444307 | 0.450827 | 0.466171 | 0.574301 | 0.611137 | 0.614950 | 0.615232 | 0.615232 | 0.615232 | malignant |
| 172.406000 | 0.442004 | 0.444307 | 0.450827 | 0.466171 | 0.574301 | 0.611137 | 0.614950 | 0.615232 | 0.615232 | 0.615232 | malignant |
[10]:
graph_partial_dependence(pipeline, X, features=('worst perimeter', 'worst radius'), grid_resolution=10)
Confusion Matrix¶
For binary or multiclass classification, we can view a confusion matrix of the classifier’s predictions. In the DataFrame output of confusion_matrix(), the column header represents the predicted labels while row header represents the actual labels.
[11]:
from evalml.model_understanding.graphs import confusion_matrix
y_pred = pipeline.predict(X)
confusion_matrix(y, y_pred)
[11]:
| benign | malignant | |
|---|---|---|
| benign | 1.000000 | 0.000000 |
| malignant | 0.009434 | 0.990566 |
[12]:
from evalml.model_understanding.graphs import graph_confusion_matrix
y_pred = pipeline.predict(X)
graph_confusion_matrix(y, y_pred)
Precision-Recall Curve¶
For binary classification, we can view the precision-recall curve of the pipeline.
[13]:
from evalml.model_understanding.graphs import graph_precision_recall_curve
# get the predicted probabilities associated with the "true" label
import woodwork as ww
y_encoded = y.ww.map({'benign': 0, 'malignant': 1})
y_pred_proba = pipeline.predict_proba(X)["malignant"]
graph_precision_recall_curve(y_encoded, y_pred_proba)
ROC Curve¶
For binary and multiclass classification, we can view the Receiver Operating Characteristic (ROC) curve of the pipeline.
[14]:
from evalml.model_understanding.graphs import graph_roc_curve
# get the predicted probabilities associated with the "malignant" label
y_pred_proba = pipeline.predict_proba(X)["malignant"]
graph_roc_curve(y_encoded, y_pred_proba)
The ROC curve can also be generated for multiclass classification problems. For multiclass problems, the graph will show a one-vs-many ROC curve for each class.
[15]:
from evalml.pipelines import MulticlassClassificationPipeline
X_multi, y_multi = evalml.demos.load_wine()
pipeline_multi = MulticlassClassificationPipeline(['Simple Imputer', 'Random Forest Classifier'])
pipeline_multi.fit(X_multi, y_multi)
y_pred_proba = pipeline_multi.predict_proba(X_multi)
graph_roc_curve(y_multi, y_pred_proba)
Binary Objective Score vs. Threshold Graph¶
Some binary classification objectives (objectives that have score_needs_proba set to False) are sensitive to a decision threshold. For those objectives, we can obtain and graph the scores for thresholds from zero to one, calculated at evenly-spaced intervals determined by steps.
[16]:
from evalml.model_understanding.graphs import binary_objective_vs_threshold
binary_objective_vs_threshold(pipeline, X, y, 'f1', steps=100)
[16]:
| threshold | score | |
|---|---|---|
| 0 | 0.00 | 0.542894 |
| 1 | 0.01 | 0.750442 |
| 2 | 0.02 | 0.815385 |
| 3 | 0.03 | 0.848000 |
| 4 | 0.04 | 0.874227 |
| ... | ... | ... |
| 96 | 0.96 | 0.854054 |
| 97 | 0.97 | 0.835165 |
| 98 | 0.98 | 0.805634 |
| 99 | 0.99 | 0.722892 |
| 100 | 1.00 | 0.000000 |
101 rows × 2 columns
[17]:
from evalml.model_understanding.graphs import graph_binary_objective_vs_threshold
graph_binary_objective_vs_threshold(pipeline, X, y, 'f1', steps=100)
Predicted Vs Actual Values Graph for Regression Problems¶
We can also create a scatterplot comparing predicted vs actual values for regression problems. We can specify an outlier_threshold to color values differently if the absolute difference between the actual and predicted values are outside of a given threshold.
[18]:
from evalml.model_understanding.graphs import graph_prediction_vs_actual
from evalml.pipelines import RegressionPipeline
X_regress, y_regress = evalml.demos.load_diabetes()
X_train, X_test, y_train, y_test = evalml.preprocessing.split_data(X_regress, y_regress, problem_type='regression')
pipeline_regress = RegressionPipeline(['One Hot Encoder', 'Linear Regressor'])
pipeline_regress.fit(X_train, y_train)
y_pred = pipeline_regress.predict(X_test)
graph_prediction_vs_actual(y_test, y_pred, outlier_threshold=50)
Now let’s train a decision tree on some data.
[19]:
pipeline_dt = BinaryClassificationPipeline(['Simple Imputer', 'Decision Tree Classifier'])
pipeline_dt.fit(X, y)
[19]:
pipeline = BinaryClassificationPipeline(component_graph=['Simple Imputer', 'Decision Tree Classifier'], parameters={'Simple Imputer':{'impute_strategy': 'most_frequent', 'fill_value': None}, 'Decision Tree Classifier':{'criterion': 'gini', 'max_features': 'auto', 'max_depth': 6, 'min_samples_split': 2, 'min_weight_fraction_leaf': 0.0}}, random_seed=0)
Tree Visualization¶
We can visualize the structure of the Decision Tree that was fit to that data, and save it if necessary.
[20]:
from evalml.model_understanding.graphs import visualize_decision_tree
visualize_decision_tree(pipeline_dt.estimator, max_depth=2, rotate=False, filled=True, filepath=None)
[20]:
Explaining Predictions¶
We can explain why the model made certain predictions with the explain_predictions function. This will use the Shapley Additive Explanations (SHAP) algorithm to identify the top features that explain the predicted value.
This function can explain both classification and regression models - all you need to do is provide the pipeline, the input features, and a list of rows corresponding to the indices of the input features you want to explain. The function will return a table that you can print summarizing the top 3 most positive and negative contributing features to the predicted value.
In the example below, we explain the prediction for the third data point in the data set. We see that the worst concave points feature increased the estimated probability that the tumor is malignant by 20% while the worst radius feature decreased the probability the tumor is malignant by 5%.
[21]:
from evalml.model_understanding.prediction_explanations import explain_predictions
table = explain_predictions(pipeline=pipeline, input_features=X, y=None, indices_to_explain=[3],
top_k_features=6, include_shap_values=True)
print(table)
Random Forest Classifier w/ Simple Imputer
{'Simple Imputer': {'impute_strategy': 'most_frequent', 'fill_value': None}, 'Random Forest Classifier': {'n_estimators': 100, 'max_depth': 6, 'n_jobs': -1}}
1 of 1
Feature Name Feature Value Contribution to Prediction SHAP Value
==============================================================================
worst concave points 0.26 ++ 0.20
mean concave points 0.11 + 0.11
mean concavity 0.24 + 0.08
worst concavity 0.69 + 0.05
worst perimeter 98.87 - -0.05
worst radius 14.91 - -0.05
The interpretation of the table is the same for regression problems - but the SHAP value now corresponds to the change in the estimated value of the dependent variable rather than a change in probability. For multiclass classification problems, a table will be output for each possible class.
Below is an example of how you would explain three predictions with explain_predictions.
[22]:
from evalml.model_understanding.prediction_explanations import explain_predictions
report = explain_predictions(pipeline=pipeline, input_features=X, y=y, indices_to_explain=[0, 4, 9], include_shap_values=True,
output_format='text')
print(report)
Random Forest Classifier w/ Simple Imputer
{'Simple Imputer': {'impute_strategy': 'most_frequent', 'fill_value': None}, 'Random Forest Classifier': {'n_estimators': 100, 'max_depth': 6, 'n_jobs': -1}}
1 of 3
Feature Name Feature Value Contribution to Prediction SHAP Value
==============================================================================
worst concave points 0.27 + 0.09
worst perimeter 184.60 + 0.09
worst radius 25.38 + 0.08
2 of 3
Feature Name Feature Value Contribution to Prediction SHAP Value
==============================================================================
worst perimeter 152.20 + 0.11
worst radius 22.54 + 0.09
worst concave points 0.16 + 0.08
3 of 3
Feature Name Feature Value Contribution to Prediction SHAP Value
==============================================================================
worst concave points 0.22 ++ 0.20
mean concave points 0.09 + 0.11
mean concavity 0.23 + 0.08
Explaining Best and Worst Predictions¶
When debugging machine learning models, it is often useful to analyze the best and worst predictions the model made. The explain_predictions_best_worst function can help us with this.
This function will display the output of explain_predictions for the best 2 and worst 2 predictions. By default, the best and worst predictions are determined by the absolute error for regression problems and cross entropy for classification problems.
We can specify our own ranking function by passing in a function to the metric parameter. This function will be called on y_true and y_pred. By convention, lower scores are better.
At the top of each table, we can see the predicted probabilities, target value, error, and row index for that prediction. For a regression problem, we would see the predicted value instead of predicted probabilities.
[23]:
from evalml.model_understanding.prediction_explanations import explain_predictions_best_worst
report = explain_predictions_best_worst(pipeline=pipeline, input_features=X, y_true=y,
include_shap_values=True, top_k_features=6, num_to_explain=2)
print(report)
Random Forest Classifier w/ Simple Imputer
{'Simple Imputer': {'impute_strategy': 'most_frequent', 'fill_value': None}, 'Random Forest Classifier': {'n_estimators': 100, 'max_depth': 6, 'n_jobs': -1}}
Best 1 of 2
Predicted Probabilities: [benign: 0.0, malignant: 1.0]
Predicted Value: malignant
Target Value: malignant
Cross Entropy: 0.0
Index ID: 168
Feature Name Feature Value Contribution to Prediction SHAP Value
==============================================================================
worst perimeter 155.30 + 0.10
worst radius 23.14 + 0.08
worst concave points 0.17 + 0.08
worst area 1660.00 + 0.06
mean concave points 0.10 + 0.05
area error 122.30 + 0.04
Best 2 of 2
Predicted Probabilities: [benign: 0.0, malignant: 1.0]
Predicted Value: malignant
Target Value: malignant
Cross Entropy: 0.0
Index ID: 564
Feature Name Feature Value Contribution to Prediction SHAP Value
==============================================================================
worst perimeter 166.10 + 0.10
worst radius 25.45 + 0.08
worst concave points 0.22 + 0.08
worst area 2027.00 + 0.06
mean concave points 0.14 + 0.05
mean concavity 0.24 + 0.05
Worst 1 of 2
Predicted Probabilities: [benign: 0.552, malignant: 0.448]
Predicted Value: benign
Target Value: malignant
Cross Entropy: 0.802
Index ID: 40
Feature Name Feature Value Contribution to Prediction SHAP Value
=============================================================================
smoothness error 0.00 + 0.04
mean texture 21.58 + 0.03
worst texture 30.25 + 0.02
worst area 787.90 + 0.02
worst radius 15.93 - -0.03
mean concave points 0.02 - -0.03
Worst 2 of 2
Predicted Probabilities: [benign: 0.788, malignant: 0.212]
Predicted Value: benign
Target Value: malignant
Cross Entropy: 1.55
Index ID: 135
Feature Name Feature Value Contribution to Prediction SHAP Value
==============================================================================
worst texture 33.37 + 0.05
mean texture 22.47 + 0.03
mean concave points 0.03 - -0.03
worst concave points 0.09 - -0.04
worst radius 14.49 - -0.05
worst perimeter 92.04 - -0.06
We use a custom metric (hinge loss) for selecting the best and worst predictions. See this example:
import numpy as np
def hinge_loss(y_true, y_pred_proba):
probabilities = np.clip(y_pred_proba.iloc[:, 1], 0.001, 0.999)
y_true[y_true == 0] = -1
return np.clip(1 - y_true * np.log(probabilities / (1 - probabilities)), a_min=0, a_max=None)
report = explain_predictions_best_worst(pipeline=pipeline, input_features=X, y_true=y,
include_shap_values=True, num_to_explain=5, metric=hinge_loss)
print(report)
Changing Output Formats¶
Instead of getting the prediction explanations as text, you can get the report as a python dictionary or pandas dataframe. All you have to do is pass output_format="dict" or output_format="dataframe" to either explain_prediction, explain_predictions, or explain_predictions_best_worst.
Single prediction as a dictionary¶
[24]:
import json
single_prediction_report = explain_predictions(pipeline=pipeline, input_features=X, indices_to_explain=[3],
y=y, top_k_features=6, include_shap_values=True,
output_format="dict")
print(json.dumps(single_prediction_report, indent=2))
{
"explanations": [
{
"explanations": [
{
"feature_names": [
"worst concave points",
"mean concave points",
"mean concavity",
"worst concavity",
"worst perimeter",
"worst radius"
],
"feature_values": [
0.2575,
0.1052,
0.2414,
0.6869,
98.87,
14.91
],
"qualitative_explanation": [
"++",
"+",
"+",
"+",
"-",
"-"
],
"quantitative_explanation": [
0.19966729417702012,
0.10648831456429969,
0.07869244977813485,
0.05150874542350735,
-0.04930428857229847,
-0.05034083333027343
],
"drill_down": {},
"class_name": "malignant"
}
]
}
]
}
Single prediction as a dataframe¶
[25]:
single_prediction_report = explain_predictions(pipeline=pipeline, input_features=X, indices_to_explain=[3],
y=y, top_k_features=6, include_shap_values=True,
output_format="dataframe")
single_prediction_report
[25]:
| feature_names | feature_values | qualitative_explanation | quantitative_explanation | class_name | prediction_number | |
|---|---|---|---|---|---|---|
| 0 | worst concave points | 0.2575 | ++ | 0.199667 | malignant | 0 |
| 1 | mean concave points | 0.1052 | + | 0.106488 | malignant | 0 |
| 2 | mean concavity | 0.2414 | + | 0.078692 | malignant | 0 |
| 3 | worst concavity | 0.6869 | + | 0.051509 | malignant | 0 |
| 4 | worst perimeter | 98.8700 | - | -0.049304 | malignant | 0 |
| 5 | worst radius | 14.9100 | - | -0.050341 | malignant | 0 |
Best and worst predictions as a dictionary¶
[26]:
report = explain_predictions_best_worst(pipeline=pipeline, input_features=X, y_true=y,
num_to_explain=1, top_k_features=6,
include_shap_values=True, output_format="dict")
print(json.dumps(report, indent=2))
{
"explanations": [
{
"rank": {
"prefix": "best",
"index": 1
},
"predicted_values": {
"probabilities": {
"benign": 0.0,
"malignant": 1.0
},
"predicted_value": "malignant",
"target_value": "malignant",
"error_name": "Cross Entropy",
"error_value": 9.95074382629983e-05,
"index_id": 168
},
"explanations": [
{
"feature_names": [
"worst perimeter",
"worst radius",
"worst concave points",
"worst area",
"mean concave points",
"area error"
],
"feature_values": [
155.3,
23.14,
0.1721,
1660.0,
0.1043,
122.3
],
"qualitative_explanation": [
"+",
"+",
"+",
"+",
"+",
"+"
],
"quantitative_explanation": [
0.09988982304983156,
0.08240174808629956,
0.07868368954615064,
0.06242860386204596,
0.051970789425386396,
0.04459155806887927
],
"drill_down": {},
"class_name": "malignant"
}
]
},
{
"rank": {
"prefix": "worst",
"index": 1
},
"predicted_values": {
"probabilities": {
"benign": 0.788,
"malignant": 0.212
},
"predicted_value": "benign",
"target_value": "malignant",
"error_name": "Cross Entropy",
"error_value": 1.5499050281608746,
"index_id": 135
},
"explanations": [
{
"feature_names": [
"worst texture",
"mean texture",
"mean concave points",
"worst concave points",
"worst radius",
"worst perimeter"
],
"feature_values": [
33.37,
22.47,
0.02704,
0.09331,
14.49,
92.04
],
"qualitative_explanation": [
"+",
"+",
"-",
"-",
"-",
"-"
],
"quantitative_explanation": [
0.05245422607466413,
0.03035933540832274,
-0.03461744299818247,
-0.04174884967530769,
-0.0491285663898271,
-0.05666940833106337
],
"drill_down": {},
"class_name": "malignant"
}
]
}
]
}
Best and worst predictions as a dataframe¶
[27]:
report = explain_predictions_best_worst(pipeline=pipeline, input_features=X, y_true=y,
num_to_explain=1, top_k_features=6,
include_shap_values=True, output_format="dataframe")
report
[27]:
| feature_names | feature_values | qualitative_explanation | quantitative_explanation | class_name | label_benign_probability | label_malignant_probability | predicted_value | target_value | error_name | error_value | index_id | rank | prefix | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | worst perimeter | 155.30000 | + | 0.099890 | malignant | 0.000 | 1.000 | malignant | malignant | Cross Entropy | 0.000100 | 168 | 1 | best |
| 1 | worst radius | 23.14000 | + | 0.082402 | malignant | 0.000 | 1.000 | malignant | malignant | Cross Entropy | 0.000100 | 168 | 1 | best |
| 2 | worst concave points | 0.17210 | + | 0.078684 | malignant | 0.000 | 1.000 | malignant | malignant | Cross Entropy | 0.000100 | 168 | 1 | best |
| 3 | worst area | 1660.00000 | + | 0.062429 | malignant | 0.000 | 1.000 | malignant | malignant | Cross Entropy | 0.000100 | 168 | 1 | best |
| 4 | mean concave points | 0.10430 | + | 0.051971 | malignant | 0.000 | 1.000 | malignant | malignant | Cross Entropy | 0.000100 | 168 | 1 | best |
| 5 | area error | 122.30000 | + | 0.044592 | malignant | 0.000 | 1.000 | malignant | malignant | Cross Entropy | 0.000100 | 168 | 1 | best |
| 6 | worst texture | 33.37000 | + | 0.052454 | malignant | 0.788 | 0.212 | benign | malignant | Cross Entropy | 1.549905 | 135 | 1 | worst |
| 7 | mean texture | 22.47000 | + | 0.030359 | malignant | 0.788 | 0.212 | benign | malignant | Cross Entropy | 1.549905 | 135 | 1 | worst |
| 8 | mean concave points | 0.02704 | - | -0.034617 | malignant | 0.788 | 0.212 | benign | malignant | Cross Entropy | 1.549905 | 135 | 1 | worst |
| 9 | worst concave points | 0.09331 | - | -0.041749 | malignant | 0.788 | 0.212 | benign | malignant | Cross Entropy | 1.549905 | 135 | 1 | worst |
| 10 | worst radius | 14.49000 | - | -0.049129 | malignant | 0.788 | 0.212 | benign | malignant | Cross Entropy | 1.549905 | 135 | 1 | worst |
| 11 | worst perimeter | 92.04000 | - | -0.056669 | malignant | 0.788 | 0.212 | benign | malignant | Cross Entropy | 1.549905 | 135 | 1 | worst |