Project:OpenMLDatamodels

Dataset

Dataset from OpenML; currently, there is no way to associate these datasets with papers, unless there is a doi match

• label: OpenML dataset name
• description: OpenML dataset with id {ID}
• instance of data set (wikidata: Q1172284)
• OpenML dataset ID: unique dataset ID from OpenML
• dataset version: version of dataset
• author name string: This input was free text and can't be associated to an ID of any kind at the moment - creators and contributors are both used here
• collection date: from freeform text field --> string is used as is
• upload date: this is an automatic timestamp
• license: license name gets matched to License items in KG like in the CRAN importer
• full work available at url: both the fields "url" and "original data url" are used for this
• default target attribute (e.g. class)
• row id attribute
• OpenML semantic tag: these were automatically tagged in OpenML; only tags in this list are considered: Agriculture, Astronomy, Chemistry, Computational Universe, Computer Systems, Culture, Demographics, Earth Science, Economics, Education, Geography, Government, Health, History, Human Activities, Images, Language, Life Science, Machine Learning, Manufacturing, Mathematics, Medicine, Meteorology, Physical Sciences, Politics, Social Media, Sociology, Statistics, Text & Literature, Transportation
• citation text: this is a Publication item; if there is a doi or an arxiv ID, the importer tries to find existing papers with that ID; else, a new paper gets created
• has feature: features and their data types, such as the feature "width" with the data type "numeric"
• number of binary features
• number of classes
• number of features
• number of instances
• number of instances with missing values
• number of missing values
• number of numeric features
• number of symbolic features
• file format: ARFF or Sparse ARFF

Publication

If no publication is found for the identifier, a publication item which just consists of the identifier without a label is created

• label: None
• description: scientific article about an OpenML dataset
• arxiv ID (if present)
• doi (if present)