Journal of Engineering, Project, and Production Management, 2026, 16(5), 2026-45
An Intelligent Prediction Model for Accounting Cost Calculation Based on Temporal Convolution and Attention Mechanisms
1
Undergraduate Student, School of Accounting, Harbin University of
Commerce, Harbin, 150000, China, E-mail: wutong521222@outlook.com
(corresponding author).
2 Undergraduate Student, Business
School, University of New South Wales, Sydney, 2033, Australia, E-mail:
wushuang846034@163.com
Project Management
Received January 13, 2026; revised March 12, 2026; accepted July 16, 2026
Available online July 25, 2026
Abstract: Accounting cost series exhibit multi-scale fluctuations and sudden disturbances, making it difficult for traditional statistical models and single deep structures to achieve stable predictive performance. To improve the accuracy and robustness of accounting cost prediction, this paper proposes an intelligent model that integrates multi-scale temporal convolution and channel-temporal collaborative attention. This model constructs stable and learnable input segments through normalization, anomaly suppression, and a bidirectional sliding window, and captures multi-granularity dependencies using parallel dilated convolution. Attention is used to dynamically weight key cost elements and key time points. Experimental results show that the proposed model outperforms the comparative models in prediction accuracy, with mean absolute error, root mean square error, and mean absolute percentage error of 0.07, 0.12, and 8.50%, respectively. In multi-product line prediction, the mean absolute error stabilizes between 0.07 and 0.09, and the root mean square error is below 0.12. In the presence of noise and missing data, the root mean square error remains between 0.08 and 0.09, demonstrating the best performance in bias control. The mean absolute error for the four prediction periods is 0.30%, and the inference latency is stable at approximately 2.0ms to 2.8ms. The proposed forecast model provides effective support for enterprise cost control and rolling budget. The research satisfies the dual requirements of forecast excellence and proven decision impact, providing a robust mechanism for converting fine-grained accounting data into actionable financial intelligence, thereby minimizing structural budget waste and maximizing the strategic utility of rolling forecasts in complex business environments.
Keywords: Accounting cost forecasting; Multi-scale temporal convolution; Attention mechanism; Intelligent financial data management; Rolling budget.
Copyright © Journal of Engineering, Project, and Production Management (EPPM-Journal).
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.
Requests for reprints and permissions at eppm.journal@gmail.com.
Citation: Wu, T. and Wu, S. (2026). An Intelligent Prediction Model for Accounting Cost Calculation Based on Temporal Convolution and Attention Mechanisms. Journal of Engineering, Project, and Production Management, 16(5), 2026-45.
