通过微聚合进行公平和私有数据预处理
摘要
1简介及相关工作
1.1光洁度校正
1.2公平与隐私
1.3捐款
2背景和定义
2.1公平性
2.2隐私
![](/cms/10.1145/3617377/asset/1546b02f-e66a-458f-a578-0be2335666f6/assets/images/medium/tkdd-2022-07-0274-algo1.jpg)
3公平MDAV说明
![](/cms/10.1145/3617377/asset/29226e61-2a7a-4866-8179-fc932fbf0d1e/assets/images/medium/tkdd-2022-07-0274-algo2.jpg)
![](/cms/10.1145/3617377/asset/ad7ebdd4-c810-414c-ac85-a466b46325bd/assets/images/medium/tkdd-2022-07-0274-algo3.jpg)
![](/cms/10.1145/3617377/asset/df12c91d-9465-4c1e-82af-ffd191c0453f/assets/images/medium/tkdd-2022-07-0274-algo4.jpg)
3.1简化示例
![](/cms/10.1145/3617377/asset/3ee08a7c-feb9-4513-8c37-dcc4221db8ea/assets/images/medium/tkdd-2022-07-0274-f01.jpg)
4实验
![](/cms/10.1145/3617377/asset/370fc3cc-23d0-4475-a988-018a55227086/assets/images/medium/tkdd-2022-07-0274-f02.jpg)
4.1公平、隐私和准确性权衡
![](/cms/10.1145/3617377/asset/8b6490e6-9b79-4acf-a99e-228631c05ec3/assets/images/medium/tkdd-2022-07-0274-f03.jpg)
![](/cms/10.1145/3617377/asset/7b9bd0f6-7d2d-480b-86dc-66037b34b778/assets/images/medium/tkdd-2022-07-0274-f04.jpg)
![](/cms/10.1145/3617377/asset/88aff1d4-57a0-42f0-af06-97eb2c725e2a/assets/images/medium/tkdd-2022-07-0274-f05.jpg)
![](/cms/10.1145/3617377/asset/1667e9fc-fe51-43bf-928b-a7c83daa7968/assets/images/medium/tkdd-2022-07-0274-f06.jpg)
![](/cms/10.1145/3617377/asset/6e8dc163-733b-4368-ab7c-09f23b335405/assets/images/medium/tkdd-2022-07-0274-f07.jpg)
![](/cms/10.1145/3617377/asset/8de03533-28eb-4218-84dd-a550bda83e2c/assets/images/medium/tkdd-2022-07-0274-f08.jpg)
![](/cms/10.1145/3617377/asset/01890cbd-5551-4c0d-8bdd-62ecd445ccd5/assets/images/medium/tkdd-2022-07-0274-f09.jpg)
![](/cms/10.1145/3617377/asset/53a2751e-f40e-44af-abb6-c2b84d7bbbe1/assets/images/medium/tkdd-2022-07-0274-f10.jpg)
4.2使用 公平 -仅用于公平性校正的MDAV
4.3与现有公平性方法的比较
4.3.1人口奇偶性比较。
4.3.2均衡赔率比较。
5限制
6结论
补充材料
下载 3.10 MB
A在线资源
A.1代码可用性
A.2数据和材料的可用性
B其他地块
![](/cms/10.1145/3617377/asset/3ecb7e84-f43d-4ebf-a4d7-d44da275d584/assets/images/medium/tkdd-2022-07-0274-f11.jpg)
![](/cms/10.1145/3617377/asset/c29e5845-be35-4678-8755-7fdc0f1c3b87/assets/images/medium/tkdd-2022-07-0274-f12.jpg)
![](/cms/10.1145/3617377/asset/449f1259-aa8a-4750-a630-6754f55197e7/assets/images/medium/tkdd-2022-07-0274-f13.jpg)
![](/cms/10.1145/3617377/asset/c7e15933-b008-47b7-ae8f-053b37c744ee/assets/images/medium/tkdd-2022-07-0274-f14.jpg)
![](/cms/10.1145/3617377/asset/a286bf02-5657-4325-a730-46943a04b4fa/assets/images/medium/tkdd-2022-07-0274-f15.jpg)
![](/cms/10.1145/3617377/asset/1f7f8bdf-939b-4163-a18e-8fc082ac5b2c/assets/images/medium/tkdd-2022-07-0274-f16.jpg)
![](/cms/10.1145/3617377/asset/bc62d4ee-eed5-48ad-8450-2d65716dcfd4/assets/images/medium/tkdd-2022-07-0274-f17.jpg)
![](/cms/10.1145/3617377/asset/30887db8-09cb-47d1-928b-48351f393c21/assets/images/medium/tkdd-2022-07-0274-f18.jpg)
工具书类
建议
编辑微聚集和加性噪声的约束 PSDML’10:关于数据挖掘和机器学习中的隐私和安全问题的国际ECML/PKDD会议记录 隐私保护数据挖掘和统计披露控制提出了几种令人不安的方法来保护受访者的隐私。 这种扰动可能会给敏感数据带来不一致性。 因此,数据编辑技术。。。