Title: An abnormal detection method of enterprise financial accounting data based on Bayesian network

Authors: Baoyuan Liu

Addresses: Department of Economics and Trade, Shijiazhuang University of Applied Technology, Shijiazhuang, 050073, China

Abstract: To improve the effectiveness of anomaly detection in enterprise financial accounting data and reduce the error probability of anomaly detection, this paper proposes a Bayesian network-based anomaly detection method for enterprise financial accounting data to ensure the accuracy and reliability of financial reports. By introducing the nearest neighbour rule and KNN algorithm to calculate the distance between different data attributes, the XGBoot algorithm is used to obtain the optimal balance point and achieve the classification of enterprise financial accounting data. According to the topological structure, the prior knowledge and accounting data characteristics are fitted, the accounting data characteristics are extracted, the interference items of abnormal features are removed by Markov blanket elimination method, the conditional probability of Bayesian network is calculated, and the data anomaly detection is realised to realise the final research. The test results indicate that the false positive rate of abnormal data detected by this method is low, and the recall rate is high, which has certain feasibility.

Keywords: Bayesian network; enterprise financial accounting data; abnormal detection; nearest neighbour rule; parameter learning.

DOI: 10.1504/IJRIS.2026.150611

International Journal of Reasoning-based Intelligent Systems, 2026 Vol.18 No.1, pp.1 - 10

Received: 30 Jun 2023
Accepted: 31 Aug 2023
Published online: 18 Dec 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article