Translating artificial intelligence into socio-economic insight: a hybrid deep learning approach to employee financial well-being

Abstract

This study aims to translate recent advancements in hybrid artificial intelligence (AI) modeling into a functional tool for assessing individual financial well-being. The objective is to develop a system that aids organizations in understanding employees’ financial stress, with broader implications for enhancing workplace productivity and societal economic resilience. A deep learning pipeline was developed to classify individuals into three financial well-being categories: Financially Secure, Moderately Stable, and Financially At-Risk. The approach utilizes a structured dataset of 20,000 Indian individuals and implements 15 advanced deep learning models, including Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Gated Recurrent Units (GRU), Bidirectional Long Short-Term Memory (BiLSTM), and Wide & Deep networks. Model performance was assessed using standard evaluation metrics, including validation accuracy and ROC-AUC scores. Among the tested models, the hybrid Wide & Deep + CNN configuration yielded the highest performance, achieving a validation accuracy of 99.44% and a perfect ROC-AUC score of 1.0000. These results validate the model’s capacity for robust classification and real-world applicability to financial profiling. This study demonstrates a practical application of AI in financial decision support systems and contributes to organizational research by offering a scalable solution to assess and mitigate employee financial stress.