Scholarlink

Predicting bank term deposit subscribers is vital for financial institutions to optimize marketing. The challenge lies in handling highly imbalanced data. We propose a solution combining SMOTE for data sampling and deep learning-based feature selection within a teacher-student framework. This approach enhances model performance by identifying key features and improving subscriber classification. The teacher-student framework automatically selects informative features, boosting predictive accuracy while reducing noise. By leveraging deep learning, our method ensures the model focuses on the most discriminative attributes, leading to better generalization. Our stack ensemble model, incorporating this method, achieved an AUC of 93.31%, demonstrating superior performance. This approach enables financial institutions to refine targeting, improve decision-making, and optimize resource allocation efficiently.

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

A Deep Learning Approach to Bank Term Deposit Prediction Using Teacher-Student Feature Selection and Stacked Models

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Meryem Saket,
  • Hayat Sahlaoui,
  • Mohamed Lkhalidi,
  • Hicham Moussaid,
  • Abdi Youssouf Hamze,
  • El Arbi Abdellaoui Alaoui,
  • Mostafa Merras

摘要

Predicting bank term deposit subscribers is vital for financial institutions to optimize marketing. The challenge lies in handling highly imbalanced data. We propose a solution combining SMOTE for data sampling and deep learning-based feature selection within a teacher-student framework. This approach enhances model performance by identifying key features and improving subscriber classification. The teacher-student framework automatically selects informative features, boosting predictive accuracy while reducing noise. By leveraging deep learning, our method ensures the model focuses on the most discriminative attributes, leading to better generalization. Our stack ensemble model, incorporating this method, achieved an AUC of 93.31%, demonstrating superior performance. This approach enables financial institutions to refine targeting, improve decision-making, and optimize resource allocation efficiently.