Scholarlink

The rapid proliferation of service providers across industries, notably in banking, has inundated customers with a plethora of options for financial services. Consequently, customer churn and engagement have emerged as paramount concerns for banks worldwide. By analyzing customer behavior, the research aims to explore the likelihood of churn. This paper proposes a method utilizing machine learning techniques, a subset of artificial intelligence, to predict customer churn in the banking sector. Our main focus was comparing classical feature selection techniques using SPSS’s PCA and novel techniques using Python. Then to assess our theory, we trained Machine Learning models based on the features selected with each method and evaluated their effectiveness based on F1 score. The study employs various classifiers, including KNN, Naive Bayes and Random Forest, achieving an F1 score of 94.78 using the latter. This finding underscores the effectiveness of leveraging machine learning algorithms to address the pressing issue of customer churn in the banking industry. By identifying key predictors of churn and deploying robust predictive models, banks can proactively mitigate churn rates and enhance customer retention strategies, thereby fostering long-term profitability and sustainability in an increasingly competitive market landscape.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Comparing Classical and Novel Techniques for Feature Selection: Toward a Machine Learning Method

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Amal Ben Abdallah,
  • Maryam Elamine

摘要

The rapid proliferation of service providers across industries, notably in banking, has inundated customers with a plethora of options for financial services. Consequently, customer churn and engagement have emerged as paramount concerns for banks worldwide. By analyzing customer behavior, the research aims to explore the likelihood of churn. This paper proposes a method utilizing machine learning techniques, a subset of artificial intelligence, to predict customer churn in the banking sector. Our main focus was comparing classical feature selection techniques using SPSS’s PCA and novel techniques using Python. Then to assess our theory, we trained Machine Learning models based on the features selected with each method and evaluated their effectiveness based on F1 score. The study employs various classifiers, including KNN, Naive Bayes and Random Forest, achieving an F1 score of 94.78 using the latter. This finding underscores the effectiveness of leveraging machine learning algorithms to address the pressing issue of customer churn in the banking industry. By identifying key predictors of churn and deploying robust predictive models, banks can proactively mitigate churn rates and enhance customer retention strategies, thereby fostering long-term profitability and sustainability in an increasingly competitive market landscape.