Donated on 11/2/2010

AutoUniv is an advanced data generator for classifications tasks. The aim is to reflect the nuances and heterogeneity of real data. Data can be generated in .csv, ARFF or C4.5 formats.

Dataset Characteristics


Subject Area


Associated Tasks


Feature Type

Categorical, Integer, Real

# Instances


# Features


Dataset Information

Additional Information

The user first creates a classification model and then generates classified examples from it. To create a model, the following are specified: the number of attributes (up to 1000) and their type (discrete or continuous), the number of classes (up to 10), the complexity of the underlying rules and the noise level. AutoUniv then produces a model through a process of constrained randomised search to satisfy the user's requirements. A model can have up to 3000 rules. Rare class models can be designed. A sequence of models can be designed to reflect concept and/or population drift. AutoUniv creates three text files for a model: a Prolog specification of the model used to generate examples (.aupl); a user-friendly statement of the classification rules in an 'if ... then' format (.aurules); a statistical summary of the main properties of the model, including its Bayes rate (.auprops).

Has Missing Values?


Variable Information

Attributes may be discrete with up to 10 values or continuous. A discrete attribute can be nominal with values v1, v2, v3 ... or integer with values 0, 1, 2 , ... .

0 citations


Ray. Hickey


By using the UCI Machine Learning Repository, you acknowledge and accept the cookies and privacy practices used by the UCI Machine Learning Repository.

Read Policy