Predictive Credit Risk Scoring Neural Network

Create an advanced neural network model for credit risk assessment that incorporates non-traditional data sources, including social media signals, alternative credit history, and macroeconomic indicators. The model must demonstrate superior predictive accuracy compared to traditional credit scoring methods, with explicit explainability features and a robust validation framework. Include detailed methodology for handling class imbalance, feature engineering, and model interpretability in financial decision-making contexts.