Audit Rule Discovery Algorithm Research Based On Algebraic And Logical Constraints

At present,while China's rapid economic development,it's also strictly maintaining the market economic order,strengthening prevention of various economic risks,attaching great importance to audit work,and actively carrying out various audit activities.Auditing has always been one of the important work contents of the society at home and abroad.As an independent economic supervision activity of the country,it plays a role of restraint and promotion to the audited units.However,with the popularity of computer and network technology,data has exploded,and large amounts of data are stored in the information systems of various units.How to "intelligently and automatically" mine potentially useful audit information from these noisy data,and it is important to assist relevant researchers in auditing.At the same time,it is also facing severe challenges,and it is even more necessary to explore new audit methods and technologies at multiple levels and angles.The main work of this thesis is to study the use of database,data mining and knowledge discovery and other technologies to organically combine with the audit business under the condition of massive data.Automatically extract the implicit rules and internal relationships between the characteristic attributes of the data from the database system,and then discover potential anomalies in the data,and actively explore the intelligent and automated auditing technology.Specifically,the main work of this article includes:1)We propose an improved algorithm based on the theory of algebraic constraints and denial constraints,which can automatically discover approximate fuzzy constraint rules between attribute columns and tuples in relational data.The found constraint rules are audit rules that have good representation and are easy to understand.They cover a wide range of common business rules and can be widely used in different scenarios.At the same time,the abnormal data set that does not satisfy the constraint rules is generated as the audit trail to assist the audit work.2)To reduce the algorithm search space and improve the operation efficiency,we design a variety of pruning rules.3)To facilitate user verification the large number of mined constraint rules,we design a novel rule scoring method for sorting rules,which takes into account two dimensional factors,namely the simplicity and coverage of the rules.4)We implemented the above two algorithms in parallel,and design experiments to evaluate the scalability of the algorithm,analyze the quality of the discovered rules,and explain its auditing significance.