Residual Self-Guided Quantum Vision Adversarial Attention Network for Predicting AI-Driven Financial Performance in Automobile Industries
Year: 2026
Type: article
Abstract: The rapid adoption of artificial intelligence (AI) in the automobile banking industry and its impact on the business processes is very fast, the impact of AI techniques in financial performance is not fully understood. The challenge here is that banks involved in the automotive sector often have fragmented disclosures of AI and inconsistent metrics, making it challenging to come up with accurate predictions of the financial impact of AI. To address this gap, they propose a model known as “Residual Self-Guided Quantum Vision Adversarial Attention Network for Predicting AI-Driven Financial Performance in Automobile Industries. To fill this gap, they propose a model called “Residual Self-Guided Quantum Vision Adversarial Attention Network for Predicting AI-Driven Financial Performance in Automobile Industries.in the automobile banking industry has transformed business processes, yet how AI techniques influence financial performance is not yet fully understood. The difficulty here lies in scattered AI disclosures and disparate metrics that hinder effective forecasts of financial results in car-related banks. In order to bridge this gap, the research introduces a model called “Residual Self-Guided Quantum Vision Adversarial Attention Network for Predicting AI-Driven Financial Performance in Automobile Industries.” The analysis relies on a dataset called “AI Techniques on Financial Performance of Automobile Industries,” which contains 131 annual report records for sixteen Indian-listed banks from 2015 to 2023. This dataset captures keyword-based AI disclosures, as well as institutional and financial indicators. Data pre-processing is done through the use of the Correlation Coefficient with Min–Max Weighted (CCMMW) method to normalize and fine-tune the input parameters, while the Masked Attention Mask Transformer (Mask2Former) is utilized for feature extraction for isolating the disclosure patterns that have the greatest influence. Following these steps, a Self-Guided Quantum Generative Adversarial Network (SG-QGAN) is constructed and then advanced into the new Residual Self-Guided Quantum Single-Head Generative Elk Herd Vision Adversarial Attention Network (RSG-QSAGE-HV2AN). This network combines SG-QGAN with a Residual Attention Single-Head Vision Transformer Network (RA-SHViT-Net) and fine-tunes its parameters with the Elk Herd Optimizer (EHO). This work aims to propose a new hybrid approach named RSG-QSAGE-HV2AN to forecast AI derived financial performance of automobile financing banks. The data set is composed of ` (2015–2023), which is divided into 80% training and 20% testing. The 5-fold cross validation is used to guarantee robustness. The accuracy of the proposed model is 99.98% (95% confidence interval [99.98 ± 0.72%]) which is much better than the conventional model (Random Forest and XGBoost). The framework is combined with CCMMW pre-processing, feature extraction using Mask2Former, SG-QGAN with EHO optimized RA-SHViT-net. The results of experiments show the higher predictive performance, interpretability (through the use of SHAP) and lower computation cost.
Source: Discover Computing
Institution: Manipal University Jaipur
Language: English
Subfield: Accounting
Domain: Social Sciences
Open Access status: gold
Estimated APC paid: $1,520
OpenAlex ID: W7133527974