2.2 Source Formats

Again, a source’s format tells what kind of data was parsed for extracting rows. To some extent, it’s related to the format of the file Chapter 3 that was used to create the source. A BigML Source can have following formats:

• Table: A source extracting rows parsing CSV or JSON data.

• ARFF: A source extracting rows parsing ARFF data.

• Excel: A source extracting data from Microsoft Excel files and similar files in binary formats.

• Image: A source containing images. See more details in section 2.4 .

• Table+Image: A composite source that extracts its rows from a CSV or JSON component, which contains cells referencing images which were also included as components of the table+image composite. See more details in (section 4.4 ).

• Mixed: A composite source that contains heterogeneous components, which have different sets of fields. For a mixed source, there is no predefined method to extract data homogeneously from them. Therefore a mixed source has no fields and can not create datasets.

• Empty: A composite source that is empty or contains only empty sources.