Assignment 2 - Logistic Regression Model Comparison

Steps

1. Import the "Customer_Churn_KNIME_1000.csv" file

2. Correct any missing values by replacing them with the Mean

3. Split the data into training (80%) and test (20%) datasets

4. Apply a Random Seed Value of 12345 to the split

5. Apply a Logistic Regression learner with a target column of "Churn". Include only the following columns:

   1. TenureMonths

   2. MonthlyCharges

   3. SupportTickets

   4. ContractType
      

6. Apply a Regression Predictor

7. Apply a scoring node

8. Apply a node to display the ROC AUC Score

9. Copy the entire workflow and paste it below the original workflow

10. In the second workflow, adjust the Logistic Regression learner with a target column of "Churn". Include all columns except CustomerID.

11. Add an annotation with a red outline. Compare the Accuracy, Cohen's Kappa, Correct Classified, Error Rate, and AUC score for both models.

Logistic Regression Model Comparison Results

First, I compared both Logistic Regression models to see which one made better predictions about customer churn. Model 1 had an accuracy of 84%, while Model 2 had an accuracy of 83%. Next, Model 1 had a Cohen's Kappa score of 0.346, compared to Model 2's score of 0.306. Also, Model 1 had an AUC score of 0.829, while Model 2 had an AUC score of 0.828. Both models performed well overall, but they had more difficulty identifying customers who would leave. Finally, based on these results, I believe Model 1 performed slightly better because it had higher accuracy, Cohen's Kappa, and AUC scores. Therefore, I would choose Model 1 to help the business predict customer churn.