Audit efficiency and effectiveness can be significantly affected by data aggregation during audit procedures. Previous studies highlight that an appropriate level of data aggregation is needed because a continuous auditing (CA) system often generates numerous alarms. To respond to this issue, this study proposes a CA system with a three-layer structure. In the first layer of the proposed system, all journal entry level transactions are classified and aggregated using defined rules; any transactions that deviate from these rules are identified as unusual transactions. The second layer detects the observations that violate controls. Analytical monitoring models are developed in the final layer to identify observations that statistically deviate from an organization’s typical business behaviors. To examine whether the proposed three-layer CA system enhances the effectiveness of a CA system in identifying financial irregularities, this study empirically tests the proposed models using real-world journal entry data from a construction company. The results indicate that the proposed framework enhances audit effectiveness and efficiency.

审计过程中的数据汇总会显着影响审计效率和有效性。以前的研究强调需要适当级别的数据聚合，因为连续审计 (CA) 系统通常会生成大量警报。针对这个问题，本研究提出了三层结构的CA系统。在提议系统的第一层，所有日记帐分录级交易都使用定义的规则进行分类和汇总；任何偏离这些规则的交易都被认定为异常交易。第二层检测违反控制的观察。在最后一层开发分析监控模型，以识别在统计上偏离组织典型业务行为的观察结果。 在识别 财务违规行为时， 本研究 使用来自一家建筑公司的真实日记帐分录数据对提议的模型进行了实证测试。结果表明，拟议的框架提高了审计的有效性和效率。