動態異常區域分析

Dynamic Abnormal Regions Analysis

廖超晧(Chao-Hao Liao)

指導教授 ： 林真如

元智大學/工程學院/工業工程與管理學系/碩士(2012年)

https://doi.org/10.6838/YZU.2012.00343

摘要

隨著檢測儀器與偵測技術進步，現今研究人員可收集到諸多隨時間變化的空
間資料，例如各地空氣汙染濃度的變化、極端溫度的改變、晶圓平面上異常電氣
特性的分佈等，皆隸屬於時空型態資料。時空型態資料複雜度高，發生異常的區
域位置、範圍大小往往隨時間改變。有效鑑別動態異常區域在諸多應用領域極為
重要，可協助研究人員分析異常現象。本論文以多維主成份分析（Multiway
Principal Component Analysis, MPCA）技術為基礎，提出鑑別平均值改變之異常時空型態資料之方法，並以K-means 分群演算法將具有相同特徵的資料分群，並以T-test 檢定各群平均值是否正常或偏移，進而判斷平面資料發生異常的時間點，並確定其異常範圍，達到鑑別動態異常區域目的。藉由模擬分析顯示，本論文之鑑別效果較單點檢測精準度高，文末以本論文所提之鑑別方法分析新墨西哥州男性甲狀腺癌異常發病率之動態區域變化。

關鍵字

多維主成份分析 ； 異常區域 ； 時空型態資料

並列摘要

With the advances in inspection instruments and detection techniques, researchers may collect many types of spatial data that change over time nowadays. For example, the change of air pollutant concentrations level or extreme temperature in each area, and the electrical characteristics on wafers all belong to spatiotemporal data. Spatiotemporal data is very complicated. The location and the coverage of abnormal areas often change over time. Effectively identifying the dynamics of abnormal areas, which is important to many application fields, can help researchers analyze irregular phenomena. This research applies Multiway Principal Component Analysis (MPCA) to identify the abnormal spatiotemporal data with mean shifts. The data with similar characteristics are clustered together by using the K-means algorithm. The T-test is then applied to test whether each cluster has mean shifts. Last, the abnormal regions and their occurrence time are determined. The simulation results show that the accuracy of the proposed method is higher than that of the individual testing. The proposed method is applied to analyze the dynamic regions with abnormal incidence rates in male thyroid cancer data in New Mexico state.

並列關鍵字

Multiway Principal Component Analysis ； abnormal areas ； spatiotemporal data

參考文獻

[1] Ackerman, M. J., "The visible human project," IEEE Transactions, 86(3) 504-511, 1998.

[2] Bartlett, M., Movellan J. and Sejnowski, T., "Face recognition by independent component analysis," IEEE Transactions, 13(6), 1450-1464, 2002.

[4] DeVantier, B. A. and Feldman, A. D., "Review of GIS applications in hydrologic modeling, " Journal of Water Resources Planning and Management, 119(2), 246-261, 1993.

[5] Higgs, G., "The role of GIS for health utilization studies: literature review," Springerlink, 9(2), 84-99, 2009.

[7] Kulldorff, M., "Prospective time periodic geographical disease surveillance using a scan statistic, " Journal of the Royal Statistical Society, Series A, 164(1), 61-72, 2001.