A Novel Neuro-Genetic Hybrid Algorithm for Enhanced Financial Risk Classification in High-Dimensional Data

Financial risk classification is a very important task in the financial industry that demands robust and efficient models for predicting and managing risks.Traditional methods face challenges with high dimensionality, imbalanced data, and subsequently encounter poor performance.This research overcomes these challenges by presenting a novel Neuro-Genetic Hybrid Model that integrates Deep Neural Networks (DNN) for feature learning and Genetic Algorithm (GA) for optimization, leveraging the complementary strengths of both methods.The key novelty lies in the synergistic combination of DNN's ability to automatically extract complex features and GA's capability to optimize the DNN architecture, hyperparameters, and feature selection, while Synthetic Minority Over-Sampling Technique (SMOTE) addresses class imbalance.This holistic approach ensures high performance across all metrics.Conventional models like Logistic Regression (LR), Random Forest (RF), and Extreme Gradient Boosting (XGBoost) suffer from limitations in handling class imbalance and optimizing model structure, resulting in poor accuracy and recall.The proposed model demonstrates superior performance, achieving accuracy values of 0.9920 and 0.9603 on the Lending Club and German Credit datasets, respectively.The findings highlight the model's effectiveness, outperforming baseline and state-of-the-art models in terms of precision, recall, and F1-score.Additionally, the model exhibits scalability, with a 1.41-fold speedup on large datasets, and robustness in handling class-imbalanced data through SMOTE and feature selection.In summary, our proposed hybrid model serves as a robust, efficient, and accurate method for financial risk classification, addressing major challenges in this domain and significantly improving efficiency over prior techniques.