An analyst working for a large retailer is given dataset containing 1,000 customer data records for analysis. The analyst is tasked to predict if customers are likely to churn (i.e., move to a competitor) or remain loyal (i.e. continue as customer of the retailer). The Data Dictionary for the dataset is as follows.

Attribute	Description	Datatype
Customer ID	Unique identifier for each customer	Polynomial
Last Transaction Date	Most recent date on which the customer purchased an item from the retailer	Date
Discount	Percentage discount given at the time of purchase (a value from 0.0 to 1.0 indicating 0-100% respectively)	Real
Metro Region	Metropolitan region in which the customer is located (East, West, North, South)	Polynomial
Churn	Indicates if a customer has churned (Churn = Yes) or remains loyal (Churn = No).	Polynomial

The analyst utilised RapidMiner to analyse the dataset and obtained the following decision tree and confusion matrix to classify customer churn. K-fold cross validation (with 10 folds) was used to evaluate and validate the model.

Accuracy: 81.80% +/- 3.85% (micro average: 81.80%)

	True No	True Yes	Class precision
Predicted No	618	89	87.41%
Predicted Yes	93	200	68.26%
Class recall	86.92%	69.20%

Based on the decision tree, explain what actions should be taken to prevent customers from churning? Write your answer for a senior business manager or similar non-technical reader.

Question 15 options:

Question 16 (4 points)

Interpret the results in the above confusion matrix. Briefly explain why there is a standard deviation of 3.85% on the accuracy evaluation of the model.

Question 16 options:

Question 17 (4 points)

Considering the accuracy of the model alone, can the analyst be confident whether or not this model is acceptable for the objective of the task? Justify your answer.

Question 17 options:

Question 18 (4 points)

Given the objective of the task, which aspect(s) of the decision tree should the analyst improve, if any? Justify your answer.