The fast progression of accounting informatization has resulted in the widespread use of cloud platforms to handle and process large-scale financial data. Nevertheless, depending on cloud environments makes accounting systems vulnerable to data tampering, unauthorized access, and loss of data integrity; thus, they are able to undermine in a significant way not only financial accuracy but also trust. To address these challenges, this research proposes a Deep Reinforcement Learning (drl)-Based Cloud Data Integrity Verification Algorithm (diva) designed to ensure secure, adaptive, and real-time verification of financial data (n = 1900) integrity in accounting informatization systems. The proposed model integrates the Adaptive Deer Hunting Optimized Double Deep Q-Network (ADHO-DDQN) model to enhance autonomous decision-making in identifying and responding to integrity breaches. The DRL agent learns optimal verification policies through continuous interaction, dynamically adjusting parameters and optimizing resources during audits. DDQN ensures stable decision-making, while ADHO adaptively refines parameter selection for efficiency. Preprocessed accounting logs enable the model to detect subtle inconsistencies or tampering patterns effectively. Principal Component Analysis (PCA), which keeps the most important patterns while lowering the number of dimensions in the data, is used. Experimental evaluations using real-world financial datasets demonstrate that ADHO-DDQN achieves superior performance with a high integrity verification accuracy (98.9%) compared to conventional static verification methods using Python 3.8.10. The findings highlight that integrating DRL into cloud data verification frameworks, also enhances adaptive learning capabilities in accounting systems. This research contributes a scalable and intelligent integrity assurance mechanism for secure and transparent accounting informatization in the digital era.
Yong Hou (2026) studied this question.