DOCSLIB.ORG

Foundations of Bayesian Learning from Synthetic Data

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Foundations of Bayesian Learning from Synthetic Data Foundations of Bayesian Learning from Synthetic Data Harrison Wilde Jack Jewson Sebastian Vollmer Chris Holmes Department of Statistics Barcelona GSE Department of Statistics, Department of Statistics University of Warwick Universitat Pompeu Fabra Mathematics Institute University of Oxford; University of Warwick The Alan Turing Institute Abstract (Dwork et al., 2006), to define working bounds on the probability that an adversary may identify whether a particular observation is present in a dataset, given There is significant growth and interest in the that they have access to all other observations in the use of synthetic data as an enabler for machine dataset. DP’s formulation is context-dependent across learning in environments where the release of the literature; here we amalgamate definitions regard- real data is restricted due to privacy or avail- ing adjacent datasets from Dwork et al. (2014); Dwork ability constraints. Despite a large number of and Lei (2009): methods for synthetic data generation, there are comparatively few results on the statisti- Definition 1 (("; δ)-differential privacy) A ran- cal properties of models learnt on synthetic domised function or algorithm K is said to be data, and fewer still for situations where a ("; δ)-differentially private if for all pairs of adjacent, researcher wishes to augment real data with equally-sized datasets D and D0 that differ in one another party’s synthesised data. We use observation and all S ⊆ Range(K), a Bayesian paradigm to characterise the up- dating of model parameters when learning Pr[K(D) 2 S] ≤ e" × Pr [K (D0) 2 S] + δ (1) in these settings, demonstrating that caution should be taken when applying conventional learning algorithms without appropriate con- Current state-of-the-art approaches involve the privati- sideration of the synthetic data generating sation of generative modelling architectures such as process and learning task at hand. Recent Generative Adversarial Networks (GANs), Variational results from general Bayesian updating sup- Autoencoders (VAEs) or Bayesian Networks. This is port a novel and robust approach to Bayesian achieved through adjustments to their learning pro- synthetic-learning founded on decision theory cesses such that their outputs fulfil a DP guarantee that outperforms standard approaches across specified at the point of training (e.g. Zhang et al., repeated experiments on supervised learning 2017; Xie et al., 2018; Jordon et al., 2018; Rosenblatt and inference problems. et al., 2020). Despite these contributions, a fundamen- tal question remains regarding how, from a statistical perspective, one should learn from privatised synthetic 1 Introduction data. Progress has been made for simple exponential family and regression models (Bernstein and Sheldon, 2018, 2019), but these model classes are of limited use Privacy enhancing technologies comprise an area of in modern machine learning applications. rapid growth (The Royal Society, 2019). An important aspect of this field concerns publishing privatised ver- We characterise this problem for the first time via an sions of datasets for learning; it is known that simply adoption of the M -open world viewpoint (Bernardo and anonymising the data is not sufficient to guarantee Smith, 2001) associated with model misspecification; individual privacy (e.g. Rocher et al., 2019). We in- unifying the privacy and synthetic data generation lit- stead adopt the Differential Privacy (DP) framework erature alongside recent results in generalised Bayesian updating (Bissiri et al., 2016) and minimum divergence Proceedings of the 24th International Conference on Artifi- inference (Jewson et al., 2018) to ask what it means cial Intelligence and Statistics (AISTATS) 2021, San Diego, to learn from synthetic data, and how can we improve California, USA. PMLR: Volume 130. Copyright 2021 by upon our inferences and predictions given that we ac- the author(s). knowledge its privatised synthetic nature? Foundations of Bayesian Learning from Synthetic Data This characterisation results in generative models that density f0(x) with respect to the Lebesque measure, d are ‘misspecificed by design’, owing to the constraints such that x1:n ∼ F0(x); we suppose xi 2 R . These imposed upon their design by requiring the fulfilment observations are held privately by a data keeper K. of a DP guarantee. This inevitably leads to discrepancy • K uses data x1:n to produce an ("; δ)-differentially between the learner’s final model and the one that they private synthetic data generating mechanism (S- would otherwise have formulated if not for this DP DGP). With a slight abuse of notation we use restriction. In real-world, finite data contexts where Gε,δ(x1:n) to denote the S-DGP, noting that Gε,δ could synthesis methods are often ‘black-box’ in nature, it is be a fully generative model, or a private release mech- difficult for a learner to fully capture and understand anism that acts directly on the finite data x1:n (see the inherent differences in the underlying distributions discussion on the details of the S-DGP below). We of the real and synthetic data that they have access to. denote the density of this S-DGP as gε,δ. • Let fθ(x) denote a learner L’s model likeli- There are two key insights that we explore in this pa- hood for F0(x), parameterised by θ with prior per following the characterisation above: Firstly, when π~(θ), and marginal (predictive) likelihood p(x) = left unchecked, the Bayesian inference machine learns R fθ(x)~π(θ)dθ. model parameters minimising the Kullback-Leibler di- θ • L’s prior may already encompass some other set of vergence (KLD) to the synthetic data generating process real-data drawn from F0 leading to π~(θ) = π(θ j (S-DGP) (Berk et al., 1966; Walker, 2013) rather than L x1:n ), for nL ≥ 0 prior observations. the true data generating process (DGP); Secondly, ro- L bust inference methods offer improved performance We adopt a decision theoretic framework (Berger, 2013), by acknowledging this misspecification, where in some in assuming that L wishes to take some optimal action cases synthetic data can otherwise significantly hinder a^ in a prediction or inference task; satisfying: learning rather than helping it. Z In order to investigate these behaviours, we experiment a^ = arg max U(x; a)F0(x)dx: (2) with models based on a mix of simulated-private and a2A real-world data to offer empirical insights on the learn- This is with respect to a user-specified utility-function ing procedure when a varying amount of real data is U(x; a) that evaluates actions in the action space A, available, and explore the optimalities in the amounts and makes precise L’s desire to learn about F0 in order of synthetic data with which to augment this real data. to accurately identify a^. The contributions of our work are summarised below: Details of the synthetic data generation mech- anism. G 1. Learning from synthetic data can lead to unpre- In defining ε,δ, we believe it is important to dictable and negative outcomes, due to varying levels differentiate between its two possible forms: of model misspecification introduced by its genera- tion and associated privacy constraints. 1. Gε,δ(x1:n) = Gε,δ(z j x1:n): G is a privacy- 2. Robust Bayesian inference offers improvements over preserving generative model fit on the real data, classical Bayes when learning from synthetic data. such as the PATE-GAN (Jordon et al., 2018), DP- 3. Real and synthetic data can be used in tandem to GAN (Xie et al., 2018) or PrivBayes (Zhang et al., achieve practical effectiveness through the discovery 2017). Privatised synthetic data is produced by in- of desirable stopping points for learning, and optimal jecting potentially heavy-tailed noise into gradient- model configurations. based learning and/or through partitioned training 4. Consideration of the preferred properties of the in- leveraging marginal distributions, aggregations and ference procedure are critical; the specific task at subsets of the data. The S-DGP provides conditional hand can determine how best to use synthetic data. independence between z1:m and x1:m and therefore no longer queries the real data after training. R 2. Gε,δ = Kε,δ(x; dz)F0(dx): A special case of this We adopt a Bayesian standpoint throughout this paper, integral comprises the convolution of F0 with some but note that many of the results also hold in the noise distribution H, such that Gε,δ = F0 ?Hε,δ. The frequentist setting. sampling distribution is therefore not a function of the private data x1:n. In this case, the number of 2 Problem Formulation samples that we can draw is limited to m ≤ n as drawing one data item requires using one sample We outline the inference problem as follows, of K’s data. Examples of this formulation include the Laplace mechanism (Dwork et al., 2014) and • Let x1:n denote a training set of n exchangeable transformation-based privatisation (Aggarwal and observations from Nature’s true DGP, F0(x) with Yu, 2004). Harrison Wilde, Jack Jewson, Sebastian Vollmer, Chris Holmes The fundamental problem of synthetic learning. (KLD) of the model from the DGP of the data (Berk L wants to learn about F0 but only has access to their et al., 1966; Walker, 2013; Bissiri et al., 2016), where R g prior π~(θ) and to z1:m ∼ Gε,δ, where Gε,δ 6≡ F0. That KLD(gε,δ k f) = log ( ε,δ=f) dGε,δ. is, the S-DGP Gε,δ(·) is ‘misspecified by design’. This As a result, if L updates their model fθ(x) using syn- claim is supported by a number of observations: thetic data z1:m ∼ Gε,δ(x1:n), then as m ! 1 they will be learning about the limiting parameter that min- • L specifies a model p(x) using beliefs about the tar- imises the KLD to the S-DGP: get F0 to be built using real data x1:n, they are then constrained by a subsequently imposed requirement KLD θG = arg min KLD (gε,δ(·) k fθ(·)) ; (3) of guaranteeing DP which instead requires consider- ε,δ θ2Θ ation of the resulting S-DGP Gε,δ; this leads to an inevitable change in their beliefs such that the re- and under regularity conditions the posterior distri- sulting model would be misspecified relative to the bution concentrates around that point, π(θ j z1:m) ! DGP F 1θKLD as m ! 1.
Load more

Recommended publications
  • Synthpop: Bespoke Creation of Synthetic Data in R
    synthpop: Bespoke Creation of Synthetic Data in R Beata Nowok Gillian M Raab Chris Dibben University of Edinburgh University of Edinburgh University of Edinburgh Abstract In many contexts, confidentiality constraints severely restrict access to unique and valuable microdata. Synthetic data which mimic the original observed data and preserve the relationships between variables but do not contain any disclosive records are one possible solution to this problem. The synthpop package for R, introduced in this paper, provides routines to generate synthetic versions of original data sets. We describe the methodology and its consequences for the data characteristics. We illustrate the package features using a survey data example. Keywords: synthetic data, disclosure control, CART, R, UK Longitudinal Studies. This introduction to the R package synthpop is a slightly amended version of Nowok B, Raab GM, Dibben C (2016). synthpop: Bespoke Creation of Synthetic Data in R. Journal of Statistical Software, 74(11), 1-26. doi:10.18637/jss.v074.i11. URL https://www.jstatsoft. org/article/view/v074i11. 1. Introduction and background 1.1. Synthetic data for disclosure control National statistical agencies and other institutions gather large amounts of information about individuals and organisations. Such data can be used to understand population processes so as to inform policy and planning. The cost of such data can be considerable, both for the collectors and the subjects who provide their data. Because of confidentiality constraints and guarantees issued to data subjects the full access to such data is often restricted to the staff of the collection agencies. Traditionally, data collectors have used anonymisation along with simple perturbation methods such as aggregation, recoding, record-swapping, suppression of sensitive values or adding random noise to prevent the identification of data subjects.
    [Show full text]
  • Synthetic Data Generation with Probabilistic Bayesian Networks
    bioRxiv preprint doi: https://doi.org/10.1101/2020.06.14.151084; this version posted September 17, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Synthetic data generation with probabilistic Bayesian Networks Grigoriy Gogoshin∗1a, Sergio Branciamore1b, and Andrei S. Rodin∗1c ∗Corresponding authors 1Department of Computational and Quantitative Medicine, Beckman Research Institute, and Diabetes and Metabolism Research Institute, City of Hope National Medical Center, 1500 East Duarte Road, Duarte, CA 91010 USA [email protected] [email protected] [email protected] Abstract Bayesian Network (BN) modeling is a prominent and increasingly popular computational sys- tems biology method. It aims to construct probabilistic networks from the large heterogeneous biological datasets that reflect the underlying networks of biological relationships. Currently, a va- riety of strategies exist for evaluating BN methodology performance, ranging from utilizing artificial benchmark datasets and models, to specialized biological benchmark datasets, to simulation stud- ies that generate synthetic data from predefined network models. The latter is arguably the most comprehensive approach; however, existing implementations are typically limited by their reliance on the SEM (structural equation modeling) framework, which includes many explicit and implicit assumptions that may be unrealistic in a typical biological data analysis scenario. In this study, we develop an alternative, purely probabilistic, simulation framework that more appropriately fits with real biological data and biological network models. In conjunction, we also expand on our current understanding of the theoretical notions of causality and dependence / conditional independence in BNs and the Markov Blankets within.
    [Show full text]
  • Fidelity and Privacy of Synthetic Medical Data Review of Methods and Experimental Results
    Fidelity and Privacy of Synthetic Medical Data Review of Methods and Experimental Results June 2021 Ofer Mendelevitch, Michael D. Lesh SM MD FACC, Keywords: synthetic data; statistical fidelity; safety; privacy; data access; data sharing; open data; metrics; EMR; EHR; clinical trials; review of methods; de-identification; re-identification; deep learning; generative models 1 Syntegra © - Fidelity and Privacy of Synthetic Medical Data Table of Contents Table of Contents 2 Abstract 3 1. Introduction 3 1.1 De-Identification and Re-Identification 3 1.2 Synthetic Data 4 2. Synthetic Data in Medicine 5 2.1 The Syntegra Synthetic Data Engine and Medical Mind 6 3. Statistical Fidelity Validation 6 3.1 Record Distance Metric 6 3.2 Visualize and Compare Datasets 7 3.3 Population Statistics 8 3.4 Single Variable (Marginal) Distributions 9 3.5 Pairwise Correlation 10 3.6 Multivariate Metrics 11 3.6.1 Predictive Model Performance 11 3.6.2 Survival Analysis 12 3.6.3 Discriminator AUC 13 3.7 Clinical Consistency Assessment 13 4. Privacy Validation 13 4.1 Disclosure Metrics 13 4.1.1 Membership Inference Test 14 4.1.2 File Membership Hypothesis Test 16 4.1.3 Attribute Inference Test 16 4.2 Copy Protection Metrics 17 4.2.1 Distance to Closest Record - DCR 17 4.2.2 Exposure 18 5. Experimental Results 19 5.1 Datasets 19 5.2 Results 20 5.2.1 Statistical Fidelity 20 5.2.2 Privacy 30 6. Discussion and Analysis 34 6.1 DIG dataset results analysis 34 6.2 NIS dataset results analysis 35 6.3 TEXAS dataset results analysis 35 6.4 BREAST results analysis 36 7.
    [Show full text]
  • Federated Principal Component Analysis
    Federated Principal Component Analysis Andreas Grammenos1,3∗ Rodrigo Mendoza-Smith2 Jon Crowcroft1,3 Cecilia Mascolo1 1Computer Lab, University of Cambridge 2Quine Technologies 3Alan Turing Institute Abstract We present a federated, asynchronous, and (ε, δ)-differentially private algorithm for PCA in the memory-limited setting. Our algorithm incrementally computes local model updates using a streaming procedure and adaptively estimates its r leading principal components when only O(dr) memory is available with d being the dimensionality of the data. We guarantee differential privacy via an input-perturbation scheme in which the covariance matrix of a dataset X ∈ Rd×n is perturbed with a non-symmetric random Gaussian matrix with variance in d 2 O n log d , thus improving upon the state-of-the-art. Furthermore, contrary to previous federated or distributed algorithms for PCA, our algorithm is also invariant to permutations in the incoming data, which provides robustness against straggler or failed nodes. Numerical simulations show that, while using limited- memory, our algorithm exhibits performance that closely matches or outperforms traditional non-federated algorithms, and in the absence of communication latency, it exhibits attractive horizontal scalability. 1 Introduction In recent years, the advent of edge computing in smartphones, IoT and cryptocurrencies has induced a paradigm shift in distributed model training and large-scale data analysis. Under this new paradigm, data is generated by commodity devices with hardware limitations and severe restrictions on data- sharing and communication, which makes the centralisation of the data extremely difficult. This has brought new computational challenges since algorithms do not only have to deal with the sheer volume of data generated by networks of devices, but also leverage the algorithm’s voracity, accuracy, and complexity with constraints on hardware capacity, data access, and device-device communication.
    [Show full text]
  • Synthetic Data Sharing and Estimation of Viable Dynamic Treatment Regimes with Observational Data
    Synthetic Data Sharing and Estimation of Viable Dynamic Treatment Regimes with Observational Data by Nina Zhou A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy (Biostatistics) in The University of Michigan 2021 Doctoral Committee: Professor Ivo D. Dinov, Co-Chair Professor Lu Wang, Co-Chair Research Associate Professor Daniel Almirall Assistant Professor Zhenke Wu Professor Chuanwu Xi Nina Zhou [email protected] ORCID iD: 0000-0002-1649-3458 c Nina Zhou 2021 ACKNOWLEDGEMENTS Firstly, I would like to express my sincere gratitude to my advisors Dr. Lu Wang and Dr. Ivo Dinov for their continuous support of my Ph.D. study and related research, for their patience, motivation, and immense knowledge. Their guidance helped me in all the time of research and writing of this thesis. Besides my advisors, I am grateful to the the rest of my thesis committee: Dr. Zhenke Wu, Dr. Daniel Almirall and Dr.Chuanwu Xi for their insightful comments and encouragements, but also for the hard question which incented me to widen my research from various perspectives. I would like to thank my fellow doctoral students for their feedback, cooperation and of course friendship. In addition, I would like to express my gratitude to Dr. Kristen Herold for proofreading all my writings. Last but not the least, I would like to thank my family: my parents and grand parents for supporting me spiritually throughout writing this thesis and my life in general. ii TABLE OF CONTENTS ACKNOWLEDGEMENTS :::::::::::::::::::::::::: ii LIST OF FIGURES ::::::::::::::::::::::::::::::: vi LIST OF TABLES :::::::::::::::::::::::::::::::: viii LIST OF ALGORITHMS ::::::::::::::::::::::::::: x ABSTRACT ::::::::::::::::::::::::::::::::::: xi CHAPTER I.
    [Show full text]
  • Benchmarking Unsupervised Outlier Detection with Realistic Synthetic Data
    Benchmarking Unsupervised Outlier Detection with Realistic Synthetic Data GEORG STEINBUSS, Karlsruhe Institute of Technology (KIT), Germany KLEMENS BÖHM, Karlsruhe Institute of Technology (KIT), Germany Benchmarking unsupervised outlier detection is difficult. Outliers are rare, and existing benchmark data contains outliers with various and unknown characteristics. Fully synthetic data usually consists of outliers and regular instance with clear characteristics and thus allows for a more meaningful evaluation of detection methods in principle. Nonetheless, there have only been few attempts to include synthetic data in benchmarks for outlier detection. This might be due to the imprecise notion of outliers or to the difficulty to arrive at a good coverage of different domains with synthetic data. In this work we propose a generic processfor the generation of data sets for such benchmarking. The core idea is to reconstruct regular instances from existing real-world benchmark data while generating outliers so that they exhibit insightful characteristics. This allows both for a good coverage of domains and for helpful interpretations of results. We also describe three instantiations of the generic process that generate outliers with specific characteristics, like local outliers. A benchmark with state-of-the-art detection methods confirms that our generic process is indeed practical. CCS Concepts: • Mathematics of computing ! Distribution functions; • Computing methodologies ! Machine learning algorithms; Anomaly detection. Additional Key Words and Phrases: Outlier Detection, Unsupervised, Benchmark, Synthetic Data. 1 INTRODUCTION Unsupervised outlier detection aims at finding data instances in an unlabeled data set that deviate from most other instances. Such outliers tend to be rare and usually exhibit unknown characteristics. This renders the evaluation of detection performance and hence comparisons of unsupervised outlier-detection methods difficult.
    [Show full text]
  • Anomaly Detection Based on Wavelet Domain GARCH Random Field Modeling Amir Noiboar and Israel Cohen, Senior Member, IEEE
    IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 45, NO. 5, MAY 2007 1361 Anomaly Detection Based on Wavelet Domain GARCH Random Field Modeling Amir Noiboar and Israel Cohen, Senior Member, IEEE Abstract—One-dimensional Generalized Autoregressive Con- Markov noise. It is also claimed that objects in imagery create a ditional Heteroscedasticity (GARCH) model is widely used for response over several scales in a multiresolution representation modeling financial time series. Extending the GARCH model to of an image, and therefore, the wavelet transform can serve as multiple dimensions yields a novel clutter model which is capable of taking into account important characteristics of a wavelet-based a means for computing a feature set for input to a detector. In multiscale feature space, namely heavy-tailed distributions and [17], a multiscale wavelet representation is utilized to capture innovations clustering as well as spatial and scale correlations. We periodical patterns of various period lengths, which often ap- show that the multidimensional GARCH model generalizes the pear in natural clutter images. In [12], the orientation and scale casual Gauss Markov random field (GMRF) model, and we de- selectivity of the wavelet transform are related to the biological velop a multiscale matched subspace detector (MSD) for detecting anomalies in GARCH clutter. Experimental results demonstrate mechanisms of the human visual system and are utilized to that by using a multiscale MSD under GARCH clutter modeling, enhance mammographic features. rather than GMRF clutter modeling, a reduced false-alarm rate Statistical models for clutter and anomalies are usually can be achieved without compromising the detection rate. related to the Gaussian distribution due to its mathematical Index Terms—Anomaly detection, Gaussian Markov tractability.
    [Show full text]
  • SYNC: a Copula Based Framework for Generating Synthetic Data from Aggregated Sources
    SYNC: A Copula based Framework for Generating Synthetic Data from Aggregated Sources Zheng Li Yue Zhao Jialin Fu Northeastern University Toronto. H. John Heinz III College University of Toronto & Arima Inc. Carnegie Mellon University Toronto, Canada Toronto, Canada Pittsburgh, USA [email protected] [email protected] [email protected] Abstract—A synthetic dataset is a data object that is generated personally-identifiable, it can be published without concerns of programmatically, and it may be valuable to creating a single releasing personal information. However, practitioners often dataset from multiple sources when direct collection is difficult find individual level data far more appealing, as aggregated or costly. Although it is a fundamental step for many data science tasks, an efficient and standard framework is absent. data lack information such as variances and distributions of In this paper, we study a specific synthetic data generation task variables. For the downscaled synthetic data to be useful, it called downscaling, a procedure to infer high-resolution, harder- needs to be fair and consistent. The first condition means to-collect information (e.g., individual level records) from many that simulated data should mimic realistic distributions and low-resolution, easy-to-collect sources, and propose a multi-stage correlations of the true population as closely as possible. The framework called SYNC (Synthetic Data Generation via Gaussian Copula). For given low-resolution datasets, the central idea of second condition implies that when we aggregate downscaled SYNC is to fit Gaussian copula models to each of the low- samples, the results need to be consistent with the original resolution datasets in order to correctly capture dependencies data.
    [Show full text]
  • Optimal Principal Component Analysis of STEM XEDS Spectrum Images Pavel Potapov1,2* and Axel Lubk2
    Potapov and Lubk Adv Struct Chem Imag (2019) 5:4 https://doi.org/10.1186/s40679-019-0066-0 RESEARCH Open Access Optimal principal component analysis of STEM XEDS spectrum images Pavel Potapov1,2* and Axel Lubk2 Abstract STEM XEDS spectrum images can be drastically denoised by application of the principal component analysis (PCA). This paper looks inside the PCA workfow step by step on an example of a complex semiconductor structure con- sisting of a number of diferent phases. Typical problems distorting the principal components decomposition are highlighted and solutions for the successful PCA are described. Particular attention is paid to the optimal truncation of principal components in the course of reconstructing denoised data. A novel accurate and robust method, which overperforms the existing truncation methods is suggested for the frst time and described in details. Keywords: PCA, Spectrum image, Reconstruction, Denoising, STEM, XEDS, EDS, EDX Background components [3–9]. In general terms, PCA reduces the Scanning transmission electron microscopy (STEM) dimensionality of a large dataset by projecting it into an delivers images of nanostructures at high spatial resolu- orthogonal basic of lower dimension. It can be shown tion matching that of broad beam transmission electron that among all possible linear projections, PCA ensures microscopy (TEM). Additionally, modern STEM instru- the smallest Euclidean diference between the initial and ments are typically equipped with electron energy-loss projected datasets or, in other words, provides the mini- spectrometers (EELS) and/or X-rays energy-dispersive mal least squares errors when approximating data with a spectroscopy (XEDS, sometimes abbreviated as EDS smaller number of variables [10].
    [Show full text]
  • Comparing Fully and Partially Synthetic Datasets for Statistical Disclosure Control in the German IAB Establishment Panel
    TRANSACTIONS ON DATA PRIVACY 1 (2008) 105 —130 Comparing Fully and Partially Synthetic Datasets for Statistical Disclosure Control in the German IAB Establishment Panel Jörg Drechsler*, Stefan Bender* and Susanne Rässler** * Institute for Employment Research (IAB) Regensburger Straße 104 90478 Nürnberg, Germany E‐mail [email protected]; [email protected] ** Otto‐Friedrich‐University Bamberg, Department of Statistics and Econometrics Feldkirchenstraße 21, 96045 Bamberg, Germany E‐mail [email protected]‐bamberg.de Abstract.1 For datasets considered for public release, statistical agencies have to face the dilemma of guaranteeing the confidentiality of survey respondents on the one hand and offering sufficiently detailed data for scientific use on the other hand. For that reason a variety of methods that address this problem can be found in the literature. In this paper we discuss the advantages and disadvantages of two approaches that pro‐ vide disclosure control by generating synthetic datasets: The first, proposed by Rubin [1], generates fully synthetic datasets while the second suggested by Little [2] imputes values only for selected variables that bear a high risk of disclosure. Changing only some variables in general will lead to higher analytical validity. However, the disclosure risk will also increase for partially synthetic data, since true values remain in the data‐ sets. Thus, agencies willing to release synthetic datasets will have to decide, which of the two methods balances best the trade‐off between data utility and disclosure risk for their data. We offer some guidelines to help making this decision. 1 The research provided in this paper is part of the project “Wirtschaftsstatistische Paneldaten und faktische Anonymisierung” financed by the Federal Ministry for Education and Research (BMBF) and conducted by: The Federal Statistical Office Germany, the Statistical Offices of the Länder, the Institute for Applied Economic Research (IAW), and the Institute for Employment Research (IAB).
    [Show full text]
  • Mtcopula: Synthetic Complex Data Generation Using Copula Fodil Benali, Damien Bodénès, Nicolas Labroche, Cyril De Runz
    MTCopula: Synthetic Complex Data Generation Using Copula Fodil Benali, Damien Bodénès, Nicolas Labroche, Cyril de Runz To cite this version: Fodil Benali, Damien Bodénès, Nicolas Labroche, Cyril de Runz. MTCopula: Synthetic Complex Data Generation Using Copula. 23rd International Workshop on Design, Optimization, Languages and Analytical Processing of Big Data (DOLAP), 2021, Nicosia, Cyprus. pp.51-60. hal-03188317 HAL Id: hal-03188317 https://hal.archives-ouvertes.fr/hal-03188317 Submitted on 1 Apr 2021 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. MTCopula: Synthetic Complex Data Generation Using Copula Fodil Benali, Damien Bodénès Nicolas Labroche, Cyril de Runz Adwanted Group BDTLN - LIFAT, University of Tours Paris, France Blois, France {fbenali,dbodenes}@adwanted.com {nicolas.labroche,cyril.derunz}@univ-tours.fr ABSTRACT Nevertheless, recently, there has been a growing interest in Nowadays, marketing strategies are data-driven, and their quality Copula-based models for estimating [1, 26] and sampling [10, 29] depends significantly on the quality and quantity of available data. from a multivariate distribution function. Copula [15] are joint As it is not always possible to access this data, there is a need for probability distributions in which any univariate continuous synthetic data generation.
    [Show full text]
  • Data Stream Clustering Techniques, Applications, and Models: Comparative Analysis and Discussion
    big data and cognitive computing Review Data Stream Clustering Techniques, Applications, and Models: Comparative Analysis and Discussion Umesh Kokate 1,*, Arvind Deshpande 1, Parikshit Mahalle 1 and Pramod Patil 2 1 Department of Computer Engineering, SKNCoE, Vadgaon, SPPU, Pune 411 007 India; [email protected] (A.D.); [email protected] (P.M.) 2 Department of Computer Engineering, D.Y. Patil CoE, Pimpri, SPPU, Pune 411 007 India; [email protected] * Correspondence: [email protected]; Tel.: +91-989-023-9995 Received: 16 July 2018; Accepted: 10 October 2018; Published: 17 October 2018 Abstract: Data growth in today’s world is exponential, many applications generate huge amount of data streams at very high speed such as smart grids, sensor networks, video surveillance, financial systems, medical science data, web click streams, network data, etc. In the case of traditional data mining, the data set is generally static in nature and available many times for processing and analysis. However, data stream mining has to satisfy constraints related to real-time response, bounded and limited memory, single-pass, and concept-drift detection. The main problem is identifying the hidden pattern and knowledge for understanding the context for identifying trends from continuous data streams. In this paper, various data stream methods and algorithms are reviewed and evaluated on standard synthetic data streams and real-life data streams. Density-micro clustering and density-grid-based clustering algorithms are discussed and comparative analysis in terms of various internal and external clustering evaluation methods is performed. It was observed that a single algorithm cannot satisfy all the performance measures.
    [Show full text]