Cognitive Banking Systems: An Artificial Intelligence Framework for Financial Decision Governance Under Market Uncertainty

Banking systems are increasingly adopting artificial intelligence (AI) techniques to improve credit‐risk management, financial decision intelligence, and stability. However, to handle dynamic market and transaction risks, these models usually need governance techniques. To overcome this limitation, this research proposes a cognitive banking system for financial decision governance via a Slime Mold Optimizer‐based Multihead Attention Long Short‐Term Memory (SMO‐MA‐LSTM) model. This model employs the publicly available Bank Customer Payment Behavior (BCPB) Kaggle dataset, which is based on product and customer attributes used to model credit risk. The dataset is preprocessed for missing values and normalization for effective learning, utilizing principal component analysis (PCA), to reduce data dimensionality while retaining the most significant features. The preprocessed data is transformed into behavioral sequences and sent into the LSTM unit, which detects temporal patterns in payment behaviors. The multihead attention mechanism increases model interpretability by dynamically focusing on important financial factors, while the SMO is used for hyperparameter optimization to improve stability and accuracy. The system facilitates dynamic risk management and enables smart decision‐making for financial risk governance and uncertainty. The experimentation through accuracy (0.97), precision (0.92), recall (0.91), and F1‐score (0.92) metrics shows the enhanced trust in prediction and stability, confirming that the proposed cognitive banking system facilitates enhanced credit risk and decision‐making in the banking sector.