摘 要 2012年底,內政部已完成全國統計區分類系統的建置,在原本鄉鎮市區空間層級之下,新劃設了最小統計區、一級發布區與二級發布區三個空間層級。政府單位將來釋出統計資料時,也增加彙整這三個空間層級的資料,期望能提供更細緻的空間精度資料。然而在採用更細緻的空間精度時,會有更高的暴露隱私風險,在相關法規規定下,一般使用者不得取得低於隱私門檻值的資料,資料發布單位必須將低於隱私門檻值的資料進行遮罩(masking),如此一來,使用者拿到的資料將是不全面的,或者資料發布單位提高資料發布的層級,使用者拿到的資料可能較為全面,但卻損失了空間單元的細緻度。本研究從「屬性完整度」和「空間準確度」兩個指標,進行最小統計區、一級發布區、二級發布區三個空間層級的比較,分別從空間分布(spatial distribution)、空間型態(spatial pattern)、屬性基本統計(basic statistics of attribute)三個面向進行分析。結果發現最小統計區雖然空間單元最細緻,但並非永遠是最佳的選擇,不同資料項目分別擁有不同的合適發布層級。
ABSTRACT By the end of 2012, the Ministry of Interior has built the system of Census Geography. There are three new spatial levels for data dissemination: the statistical area, the 1st level dissemination area and the 2nd level dissemination area. Social economic statistical data can be released by these three additional levels to improve its accuracy. However, privacy exposure is the critical issue that has to be concerned. This study proposes a method to determine the level of dissemination areas in order to provide the most accurate data and consider privacy protection as well. Two indicators, attribute integrity and spatial accuracy, are employed in the study. It is found that the selection of the most appropriate dissemination area depends on the characteristics of data including spatial distribution and attribute values. It is also found that higher level dissemination area may still preserve the same spatial accuracy as the lower level areas. That is to say it is possible to consider data accuracy and privacy protection simultaneously.