【學術演講公告】112年5月9(二)
- 日期 : 2023-04-14
- 分類 : 演講公告
- 點閱 : 477
時 間：112年5月9日（星期二下午14:10~15:00）
地 點：管理學院新大樓M240
主講人：廖揚立 教授 (逢甲大學 統計學系)
講 題：Methods for Monitoring Benford Processes
Abstract
Statistical process monitoring (SPM) involves toolsets for effectively and timely detecting changes in processes. In industry 4.0, monitoring data quality is important due to the explosion of data available. Benford’s Law is one of many first digits distributions, and has been widely used as a foundation for quality assessment in data auditing. Further, the generalized Benford’s Law is a single parameter distribution that yields Benford’s Law as a special case, and therefore, permits the development of SPM strategies to detect deviations from Benford’s Law over time. Current methods to detect deviations from Benford’s Law use aggregated data sets, thus, limiting their use as online monitoring alternatives. In this paper we propose CUSUM and EWMA control charts useful in monitoring for unusual deviations from Benford’s Law. Unlike current methods, our monitoring strategies are more appropriate for online monitoring applications. The performance of the proposed CUSUM and EWMA charts are compared via the average run length (ARL). The results suggest the CUSUM charts have better relative ARL performances when the magnitude of shift is prespecified and known; otherwise, the EWMA charts may be considered. Application of the proposed methods are demonstrated on two open-source data sets, daily transactions and transfers of 0x (ZRX) cryptocurrency, and daily market value changes of Apple Inc (APPL).
