Improving bank customer churn prediction with feature reduction using GA.

Area of Science:

• Banking and Financial Services
• Data Science and Machine Learning
• Computational Intelligence

Background:

• Customer churn is a significant challenge for banks due to increased competition.
• High-dimensional data in customer churn datasets negatively impacts machine learning algorithm performance.
• Effective churn prediction is crucial for banks to maintain customer loyalty and plan services.

Purpose of the Study:

• To enhance the prediction accuracy of machine learning models for high-dimensional customer churn data.
• To implement feature selection using a Genetic Algorithm (GA) to improve model performance.
• To identify critical features influencing customer churn for strategic banking decisions.

Main Methods:

• Feature selection using a Genetic Algorithm (GA) was applied to high-dimensional customer churn datasets.
• The study evaluated the sensitivity of different machine learning algorithms to feature selection.
• GA-optimized classification models were developed and tested on a primary dataset.

Main Results:

• The Genetic Algorithm (GA) effectively improved prediction accuracy for high-dimensional customer churn data.
• The research identified specific features that are most important for predicting customer churn using GA-optimized models.
• The study differentiated between machine learning algorithms sensitive and insensitive to feature selection.

Conclusions:

• Feature selection with Genetic Algorithms (GA) is a viable method to enhance customer churn prediction accuracy in banking.
• Identifying key churn predictors enables banks to refine marketing strategies and reduce customer attrition.
• The approach assists financial institutions in optimizing customer retention efforts through data-driven insights.