Each dataset (containing one or more files) is uploadable through the project page.
Files must be downloaded by pairs, one file for train, one file for test. The files have to be formatted the same way.
N.B.: Files have to be encoded in UTF-8 or ASCII.
In a project, the settings are the same in all files.
The field separator can be chosen amongst pipe (|), coma (,), semicolon (;), or tabulation (\t).
Each entity of the training dataset has to be identified by a unique key. The name of this key field has to be specified to PredicSis.ai
In the central files, the outcome feature, for each entity, has to be present and specified to PredicSis.ai.
N.B.: As the key and outcome features has to be present in the central files, each file should at least contain 2 columns.
The report, and later the scoring, will be dependant of the user’s choice of the main modality. This main modality has, of course, to be present in the outcome column.
N.B.: At least 2 modalities have to be present in the outcome field. This means that each file should contain at least 3 lines, one for the header, and two for two distinct modalities.
Headers are compulsory in all files, to identify the features,
- Fields can be quoted by double quote (") as long as it does not contain a carriage return,
- Headers should not contain back ticks (`).