Moving from classification to regression in SAP Analytics Cloud — building a credit-score prediction model on the Bank Customer Churn dataset with 96.7% prediction confidence.
Building on SAC's smart prediction features, this moves from classification to regression analysis, using the Bank Customer Churn dataset to demonstrate real-world applications.
What is regression?
Regression analysis models numerical data with a continuous target variable and predictor variables, finding the 'best fit' by estimating regression parameters. The target must be continuous rather than categorical.
Business scenario
Potential continuous targets include Credit Score (300–850), Account Balance, Estimated Salary, Age and Tenure. The use case predicts credit scores using features such as country, gender, age, tenure, balance, product count and salary — with a critical note to exclude unique identifiers like customer ID, which lack predictive value.
Model development and results
The regression model is guided by Root Mean Squared Error (RMSE) analysis — lower RMSE indicates a better fit between predicted and actual values. After testing multiple models, Model 5 was selected, achieving 96.714% prediction confidence.
Conclusion
This demonstrates SAC's effectiveness for complex predictive scenarios, emphasising the importance of feature engineering — and previewing time-series analysis as the next step.
