churneval 1.1

churneval is a python package for evaluating churn models

Package "churneval"

Version: 1.1

Author: Soumi De

Maintained by: Soumi De <[e-mail]>

Description:

churneval is a package to evaluate models used in churn classification. The evaluation metrics include accuracy, sensitivity, specificity, precision, F1-score and top-decile lift.

License: GPL-3

Date: 12th November, 2021

---

Function:

get_performance_metrics               Function that returns evaluation metrics

---

Usage:

from churneval import get_performance_metrics

get_performance_metrics(model_name, true_class, predicted_class, predicted_probs)

Arguments:

• model_name: Abbreviated name of the churn model (in text)
• true_class: A dataframe of true class labels with shape (n,1)
• predicted_class: An array of binary predicted class with shape (n,)
• predicted_probs: An array of predicted class probabilities with shape (n,)

Returned Values:

A dataframe consisting of elements given below:

• Model_Name: Abbreviated name of the churn model
• Accuracy: Accuracy of churn model
• Confusion Matrix: A 2X2 array representing confusion matrix
• Precision: Precision value
• Sensitivity: Sensitivity value
• Specificity: Specificity value
• F1-score: F1-score
• ROC_score: Area under the curve
• top_dec_lift: Top decile lift value

---

Function:

top_decile_lift               Function that returns top decile lift of a sample

---

Usage:

from churneval import top_decile_lift

top_decile_lift(true_class, predicted_class, predicted_probs)

Arguments:

• true_class: A dataframe of true class labels with shape (n,1)
• predicted_class: An array of binary predicted class with shape (n,)
• predicted_probs: An array of predicted class probabilities with shape (n,)

Returned Values:

• A float object with top decile lift value

---

Function:

lift_curve               Function that plots lift curve of a model

---

Usage:

from churneval import lift_curve

lift_curve(true_class, predicted_class, predicted_probs)

Arguments:

• true_class: A dataframe of true class labels with shape (n,1)
• predicted_class: An array of binary predicted class with shape (n,)
• predicted_probs: An array of predicted class probabilities with shape (n,)

Returned Values:

• A plot that shows lift curve
  • x-axis: Proportion of data
  • y-axis: Lift of the model