Dirty data

Abstract: Detecting and repairing dirty data is one of the perennial challenges in data analytics, and failure to do so can result in inaccurate analytics and unreliable decisions. Over the past few years, there has been a surge of interest from both industry and academia on data cleaning problems including new abstractions, interfaces, approaches for scalability, and statistical techniques. To better understand the new advances in the field, we will first present a taxonomy of the data cleaning literature in which we highlight the recent interest in techniques that use constraints, rules, or patterns to detect errors, which we call qualitative data cleaning. We will describe the state-of-the-art techniques and also highlight their limitations with a series of illustrative examples. While traditionally such approaches are distinct from quantitative approaches such as outlier detection, we also discuss recent work that casts such approaches into a statistical estimation framework including: using Machine Learning to improve the efficiency and accuracy of data cleaning and considering the effects of data cleaning on statistical analysis.

Abstract: Law enforcement agencies are increasingly using predictive policing systems to forecast criminal activity and allocate police resources. Yet in numerous jurisdictions, these systems are built on data produced during documented periods of flawed, racially biased, and sometimes unlawful practices and policies (“dirty policing”). These policing practices and policies shape the environment and the methodology by which data is created, which raises the risk of creating inaccurate, skewed, or systemically biased data (“dirty data”). If predictive policing systems are informed by such data, they cannot escape the legacies of the unlawful or biased policing practices that they are built on. Nor do current claims by predictive policing vendors provide sufficient assurances that their systems adequately mitigate or segregate this data. In our research, we analyze thirteen jurisdictions that have used or developed predictive policing tools while under government commission investigations or federal court monitored settlements, consent decrees, or memoranda of agreement stemming from corrupt, racially biased, or otherwise illegal policing practices. In particular, we examine the link between unlawful and biased police practices and the data available to train or implement these systems. We highlight three case studies: (1) Chicago, an example of where dirty data was ingested directly into the city’s predictive system; (2) New Orleans, an example where the extensive evidence of dirty policing practices and recent litigation suggests an extremely high risk that dirty data was or could be used in predictive policing; and (3) Maricopa County, where despite extensive evidence of dirty policing practices, a lack of public transparency about the details of various predictive policing systems restricts a proper assessment of the risks. The implications of these findings have widespread ramifications for predictive policing writ large. Deploying predictive policing systems in jurisdictions with extensive histories of unlawful police practices presents elevated risks that dirty data will lead to flawed or unlawful predictions, which in turn risk perpetuating additional harm via feedback loops throughout the criminal justice system. The use of predictive policing must be treated with high levels of caution and mechanisms for the public to know, assess, and reject such systems are imperative.

Abstract: Analysts often clean dirty data iteratively--cleaning some data, executing the analysis, and then cleaning more data based on the results. We explore the iterative cleaning process in the context of statistical model training, which is an increasingly popular form of data analytics. We propose ActiveClean, which allows for progressive and iterative cleaning in statistical modeling problems while preserving convergence guarantees. ActiveClean supports an important class of models called convex loss models (e.g., linear regression and SVMs), and prioritizes cleaning those records likely to affect the results. We evaluate ActiveClean on five real-world datasets UCI Adult, UCI EEG, MNIST, IMDB, and Dollars For Docs with both real and synthetic errors. The results show that our proposed optimizations can improve model accuracy by up-to 2.5x for the same amount of data cleaned. Furthermore for a fixed cleaning budget and on all real dirty datasets, ActiveClean returns more accurate models than uniform sampling and Active Learning.

Abstract: Correctly detecting the semantic type of data columns is crucial for data science tasks such as automated data cleaning, schema matching, and data discovery. Existing data preparation and analysis systems rely on dictionary lookups and regular expression matching to detect semantic types. However, these matching-based approaches often are not robust to dirty data and only detect a limited number of types. We introduce Sherlock, a multi-input deep neural network for detecting semantic types. We train Sherlock on $686,765$ data columns retrieved from the VizNet corpus by matching $78$ semantic types from DBpedia to column headers. We characterize each matched column with $1,588$ features describing the statistical properties, character distributions, word embeddings, and paragraph vectors of column values. Sherlock achieves a support-weighted F$_1$ score of $0.89$, exceeding that of machine learning baselines, dictionary and regular expression benchmarks, and the consensus of crowdsourced annotations.