@techreport{oai:ipsj.ixsq.nii.ac.jp:00197695,
 author = {戸﨑, 祐輔 and 鈴木, 孝彦 and 峯, 恒憲 and 廣川, 佐千男 and Yusuke, Tozaki and Takahiko, Suzuki and Tsunenori, Mine and Sachio, Hirokawa},
 issue = {33},
 month = {Jun},
 note = {多くの数値データについて，最上位桁の数字の出現確率に法則性があり，ベンフォードの法則として知られいる．この法則は統計データ不正検出に使われている．2018年，厚生労働省が公表している障害者雇用状況の集計結果について誤りが判明し，修正が行われた．本論文では，まず，修正前後において，集計結果がベンフォードの法則に従うか否かを調べ，ベンフォードの法則の有用性を確かめる．さらに，複数の k 進法 （k=3，4，…） 上のベンフォードの法則を用いて，数値データの誤り箇所を推定する手法提案する．障害者雇用状況の集計結果をいて，推定性能を評価する．, For wide variety of numerical data, there is a rule in the distribution of the first significant digit, which is known as Benford's law. This rule is used to detect fraudulent statistics. In 2018, many errors were reported in "Employment statistics of disabled persons in Japan" published by Japan Ministry of Health, Labor and Welfare. After that, the revised statistics were published. In this paper, we firstly confirm whether the original and the revised statistics follow Benford's law or not. Following to that, we propose a new method that utilizes multiple views of Benford's law in k-adic system (k = 3, 4,...). The proposed method can detect the numbers which are candidates of correction in statistics beforehand. We evaluate the performance of the proposed method by using the original and the revised employment statistics.},
 title = {ベンフォードの法則による障害者雇用状況集計結果の誤り箇所推定},
 year = {2019}
}