A probabilistic database approach to autoencoder-based data cleaning

TL;DR

This paper introduces a probabilistic database-based autoencoder method for automatic data cleaning that learns data structure to identify and correct errors without needing clean data, improving data quality in categorical and numeric datasets.

Contribution

It presents a novel autoencoder approach combined with probabilistic databases for data cleaning, capable of handling noisy data without prior clean datasets.

Findings

Effectively removes noise from probabilistic data

Improves data quality with minimal manual cleaning

Applicable to both categorical and numeric data

Abstract

Data quality problems are a large threat in data science. In this paper, we propose a data-cleaning autoencoder capable of near-automatic data quality improvement. It learns the structure and dependencies in the data and uses it as evidence to identify and correct doubtful values. We apply a probabilistic database approach to represent weak and strong evidence for attribute value repairs. A theoretical framework is provided, and experiments show that it can remove significant amounts of noise (i.e., data quality problems) from categorical and numeric probabilistic data. Our method does not require clean data. We do, however, show that manually cleaning a small fraction of the data significantly improves performance.