BayesWipe is a product of the information integration
research group of Arizona State University
Unlike other data cleaning tools, BayesWipe is built to learn both the generative and error model directly from the dirty data itself. It uses a Bayesian network to model the generative model of the data, and a maximum entropy model (with edit distance and a substitution probability measure) for the error model.
The tool will first read the input data, strip it away of all unique keys, quantize any continuous attributes, and then train a Bayesian model on the resulting data. This Bayesian learner ignores small perturbations, and learns the overall model of the data. It also then estimates error statistics on this data. Then it processes the input data tuple by tuple, and cleans them.
In order to clean a tuple, BayesWipe first creates a set of candidate clean tuples as possible replacements for the tuple. The original tuple is always made a part of this set. Suppose the candidate tuple is T*, and the original tuple is T, BayesWipe finds that particular T* that gives the maximum probability for P(T*|T).
Further details can be found in our paper.
The BayesWipe binary is available under the BSD 2-clause license
Copyright (c) 2014, Arizona State University. All rights reserved.
Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
The BayesWipe logo uses the database icon licensed under Creative Commons – Attribution (CC BY 3.0) "Database" designed by Dmitry Baranovskiy from the Noun Project.