R包synthpop的Python实现
py-synthpop的Python项目详细描述
合成器
R package synthpop的Python实现。在
synthpop的R实现是一个工具，用于生成包含机密信息的微数据的合成版本，以便安全地发布给用户进行探索性分析。生成合成数据的关键目标是用合成的原始值替换敏感的原始值，从而使数据集中包含的统计信息失真最小。变量可以是分类的，也可以是连续的，使用序列建模方法逐一综合。替换是通过使用参数化或分类和回归树模型从与原始数据拟合的条件分布中提取来生成的。在
{{1}在cd1>之后，通过cd1}方法生成的原始数据。如果使用默认设置，或者使用用户定义的方法，这个过程可以很大程度上实现自动化。可选参数可以用来影响披露风险和合成数据的分析质量。在
发展现状和路线图
此项目处于Alpha状态，可以在here找到路线图。在
安装
pip
pip install py-synthpop
来源
^{pr2}$示例
成人数据集
我们将使用US adult census dataset，这是从美国人口普查局数据库中提取的免费开放数据集。数据集最初是为二元分类问题设计的，任务是预测一个人的年收入是否超过5万美元。数据集是离散和连续特征的混合体，包括年龄、工作状态（工人阶级）、教育程度、婚姻状况、种族、性别、关系和每周工作时间。在
In[1]:fromdatasets.adultimportdfIn[2]:df.head()Out[2]:ageworkclassfnlwgteducationeducational-nummarital-statusoccupationrelationshipracegendercapital-gaincapital-losshours-per-weeknative-countryincome039State-gov77516Bachelors13Never-marriedAdm-clericalNot-in-familyWhiteMale2174040United-States<=50K150Self-emp-not-inc83311Bachelors13Married-civ-spouseExec-managerialHusbandWhiteMale0013United-States<=50K238Private215646HS-grad9DivorcedHandlers-cleanersNot-in-familyWhiteMale0040United-States<=50K353Private23472111th7Married-civ-spouseHandlers-cleanersHusbandBlackMale0040United-States<=50K428Private338409Bachelors13Married-civ-spouseProf-specialtyWifeBlackFemale0040Cuba<=50K
合成器
- 对成人数据集使用默认参数：
In[1]:fromsynthpopimportSynthpopIn[2]:fromdatasets.adultimportdf,dtypesIn[3]:spop=Synthpop()In[4]:spop.fit(df,dtypes)train_agetrain_workclasstrain_fnlwgttrain_educationtrain_educational-numtrain_marital-statustrain_occupationtrain_relationshiptrain_racetrain_gendertrain_capital-gaintrain_capital-losstrain_hours-per-weektrain_native-countrytrain_incomeIn[5]:synth_df=spop.generate(len(df))generate_agegenerate_workclassgenerate_fnlwgtgenerate_educationgenerate_educational-numgenerate_marital-statusgenerate_occupationgenerate_relationshipgenerate_racegenerate_gendergenerate_capital-gaingenerate_capital-lossgenerate_hours-per-weekgenerate_native-countrygenerate_incomeIn[6]:synth_df.head()Out[6]:ageworkclassfnlwgteducationeducational-nummarital-statusoccupationrelationshipracegendercapital-gaincapital-losshours-per-weeknative-countryincome021?21305511th7Never-married?Not-in-familyOtherFemale0030United-States<=50K123Private150683HS-grad9Never-marriedAdm-clericalNot-in-familyWhiteFemale0040United-States<=50K261Private19141710th6WidowedSalesNot-in-familyBlackFemale0032United-States<=50K350Private190762HS-grad9DivorcedSalesNot-in-familyWhiteMale0060United-States<=50K442Local-gov255675HS-grad9Married-civ-spouseOther-serviceHusbandBlackMale0040United-States<=50KIn[7]:spop.methodOut[7]:agesampleworkclasscartfnlwgtcarteducationcarteducational-numcartmarital-statuscartoccupationcartrelationshipcartracecartgendercartcapital-gaincartcapital-losscarthours-per-weekcartnative-countrycartincomecartdtype:objectIn[8]:spop.visit_sequenceOut[8]:age0workclass1fnlwgt2education3educational-num4marital-status5occupation6relationship7race8gender9capital-gain10capital-loss11hours-per-week12native-country13income14dtype:int64In[9]:spop.predictor_matrixOut[9]:ageworkclassfnlwgteducationeducational-nummarital-statusoccupationrelationshipracegendercapital-gaincapital-losshours-per-weeknative-countryincomeage000000000000000workclass100000000000000fnlwgt110000000000000education111000000000000educational-num111100000000000marital-status111110000000000occupation111111000000000relationship111111100000000race111111110000000gender111111111000000capital-gain111111111100000capital-loss111111111110000hours-per-week111111111111000native-country111111111111100income111111111111110
- 定义成人数据集的访问顺序：
In[1]:fromsynthpopimportSynthpopIn[2]:fromdatasets.adultimportdf,dtypesIn[3]:spop=Synthpop(visit_sequence=[0,1,5,3,2])In[4]:spop.fit(df,dtypes)train_agetrain_workclasstrain_marital-statustrain_educationtrain_fnlwgtIn[5]:synth_df=spop.generate(len(df))generate_agegenerate_workclassgenerate_marital-statusgenerate_educationgenerate_fnlwgtIn[6]:synth_df.head()Out[6]:ageworkclassfnlwgteducationmarital-status057Self-emp-not-inc327901Prof-schoolMarried-civ-spouse124Private34568Assoc-vocNever-married250Private256861HS-gradMarried-civ-spouse328Private186239Some-collegeNever-married438Private216129BachelorsDivorcedIn[7]:spop.methodOut[7]:agesampleworkclasscartfnlwgtcarteducationcarteducational-numcartmarital-statuscartoccupationcartrelationshipcartracecartgendercartcapital-gaincartcapital-losscarthours-per-weekcartnative-countrycartincomecartdtype:objectIn[8]:spop.visit_sequenceOut[8]:age0workclass1fnlwgt4education3marital-status2dtype:int64In[9]:spop.predictor_matrixOut[9]:ageworkclassfnlwgteducationmarital-statusage00000workclass10000fnlwgt11011education11001marital-status11000
许可证
这个项目正在Hazy Limited开发，并在MIT license下发布。在
- 项目
标签：
欢迎加入QQ群-->： 979659372
