• R Madhavi
• Mr B Siva Kumar Reddy

Customer churn prediction is vital in the banking sector, as retaining existing clients is more cost-effective than acquiring new ones. Traditional models like Logistic Regression and Random Forest often fail to capture complex nonlinear relationships in customer data. To address this, a Transformer-Based Deep Neural Network (TDNN) with a multi-head self-attention mechanism is proposed for efficient churn prediction. The model learns global feature interactions across attributes such as demographics, account balance, credit score, and engagement metrics. SMOTE is applied to handle class imbalance, and data preprocessing includes standardization and label encoding. Experimental results show that the proposed TDNN outperforms conventional models in accuracy, precision, recall, F1-score, and AUC. The attention mechanism enhances both interpretability and predictive power, making this approach an effective and scalable solution for proactive customer retention in banking.

TRANSFORMER-BASED DEEP NEURAL NETWORK WITH ATTENTION MECHANISM FOR CUSTOMER CHURN PREDICTION IN THE BANKING SECTOR

"TRANSFORMER-BASED DEEP NEURAL NETWORK WITH ATTENTION MECHANISM FOR CUSTOMER CHURN PREDICTION IN THE BANKING SECTOR", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.12, Issue 11, page no.b688-b697, November-2025, Available :http://www.jetir.org/papers/JETIR2511186.pdf

"TRANSFORMER-BASED DEEP NEURAL NETWORK WITH ATTENTION MECHANISM FOR CUSTOMER CHURN PREDICTION IN THE BANKING SECTOR", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.12, Issue 11, page no. ppb688-b697, November-2025, Available at : http://www.jetir.org/papers/JETIR2511186.pdf