Е-Публикации
Технически университет - София

Abstract: Globalization has stimulated the opening of the market and the accumulation of huge amounts of data, which consequently leads to an increase in the importance of the control of customs operations. However, customs data is highly imbalanced and this poses challenges in its integration and processing. Therefore, it is of prime importance to find automatic computationally intelligent solutions for customs management. The purpose of the research presented in this paper is to propose a computer-aided system for customs fraud analytics based on artificial intelligence techniques, ensuring the application and verification of methods and algorithms for integration, management, analysis and visualization of data on customs violations. The architecture of the customs violation data analysis system consists of the following components: data sources, data storage, data integration and preprocessing, real-time data flow, modeling, analysis and storage of analytical data, and visualization of the results. A machine learning approach for detecting customs fraud through unstructured data analysis is proposed. An artificial neural network designed for data analysis is designed, and the input data is divided into training data and testing data. A reduced set of statistical records related to the analysis of heterogeneous databases of different institutions, which is stored in a data warehouse, are used as experimental data. The first 80% of the data are used to train the neural network and the remaining 20% to test the trained network. Experimental results show that the calculated accuracy increases with increasing epochs and is higher for the training data and lower for the validation test data. Thus, the trained model can be saved and used to monitor for anomalies. The trained model is applied to the system to calculate new input parameters that were not used in either training or validation.

References

1. Dangsawang, B., & Nuchitprasitchai, S. (2024). A machine learning approach for detecting customs fraud through unstructured data analysis in social media. Decision Analytics Journal, 10, 100408.
2. Digiampietri, L. A., Roman, N. T., Meira, L. A., Jambeiro Filho, J., Ferreira, C. D., Kondo, A. A., & Goldenstein, S. (2008, May). Uses of artificial intelligence in the Brazilian customs fraud detection system. In DG. O (pp. 181-187).
3. Jaccard, N., Rogers, T. W., Morton, E. J., & Griffin, L. D. (2016, November). Automated detection of smuggled high-risk security threats using deep learning. In 7th International Conference on Imaging for Crime Detection and Prevention (ICDP 2016) (pp. 1-6). IET
4. Jeong, C., Kim, S., Park, J., & Choi, Y. (2022). Customs import declaration datasets. arXiv preprint arXiv:2208.02484.
5. Kim, S., Mai, T. D., Han, S., Park, S., Nguyen, D. T., So, J., & Cha, M. (2022). Active learning for human-inthe-loop customs inspection. IEEE Transactions on Knowledge and Data Engineering, 35(12), 12039- 12052.
6. Kim, S., Song, S. K., Cho, M., & Shin, S. H. (2021). Transaction Pattern Discrimination of Malicious Supply Chain using Tariff-Structured Big Data. The Journal of the Korea Contents Association, 21(2), 121- 129.
7. Kim, S., Tsai, Y. C., Singh, K., Choi, Y., Ibok, E., Li, C. T., & Cha, M. (2020, August). DATE: Dual attentive tree-aware embedding for customs fraud detection. In Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (pp. 2880-2890).

Issue