Exploratory Data Mining and Data Cleaning

ISBN-13: 9780471268512 / Angielski / Twarda / 2003 / 203 str.

Tamraparni Dasu; Theodore Johnson

Opis

• Written for practitioners of data mining, data cleaning and database management.
• Presents a technical treatment of data quality including process, metrics, tools and algorithms.
• Focuses on developing an evolving modeling strategy through an iterative data exploration loop and incorporation of domain knowledge.
• Addresses methods of detecting, quantifying and correcting data quality issues that can have a significant impact on findings and decisions, using commercially available tools as well as new algorithmic approaches.
• Uses case studies to illustrate applications in real life scenarios.
• Highlights new approaches and methodologies, such as the DataSphere space partitioning and summary based analysis techniques.

Exploratory Data Mining and Data Cleaning will serve as an important reference for serious data analysts who need to analyze large amounts of unfamiliar data, managers of operations databases, and students in undergraduate or graduate level courses dealing with large scale data analys is and data mining.

Recenzje

"Statisticians not conversant with today′s statistical take on DQ should read this book…and be stimulated to do important research in DQ." ( Journal of the American Statistical Association, March 2006)

"…uniquely integrates several approaches for data cleaning and exploration…" (Journal of Statistical Computation & Simulation, April 2004)

"...provides a uniquely integrated approach...for serious data analysts everywhere..." (Zentralblatt Math, Vol. 1027, 2004)

Spis treści

0.1 Preface.

1 Exploratory Data Mining and Data Cleaning: An Overview.

1.1 Introduction.

1.2 Cautionary Tales.

1.3 Taming the Data.

1.4 Challenges.

1.5 Methods.

1.6 EDM.

1.6.1 EDM Summaries – Parametric.

1.6.2 EDM Summaries – Nonparametric.

1.7 End­to­End Data Quality (DQ).

1.7.1 DQ in Data Preparation.

1.7.2 EDM and Data Glitches.

1.7.3 Tools for DQ.

1.7.4 End­to­End DQ: The Data Quality Continuum.

1.7.5 Measuring Data Quality.

1.8 Conclusion.

2 Exploratory Data Mining.

2.1 Introduction.

2.2 Uncertainty.

2.2.1 Annotated Bibliography.

2.3 EDM: Exploratory Data Mining.

2.4 EDM Summaries.

2.4.1 Typical Values.

2.4.2 Attribute Variation.

2.4.3 Example.

2.4.4 Attribute Relationships.

2.4.5 Annotated Bibliography.

2.5 What Makes a Summary Useful?

2.5.1 Statistical Properties.

2.5.2 Computational Criteria.

2.5.3 Annotated Bibliography.

2.6 Data­Driven Approach – Nonparametric Analysis.

2.6.1 The Joy of Counting.

2.6.2 Empirical Cumulative Distribution Function (ECDF).

2.6.3 Univariate Histograms.

2.6.4 Annotated Bibliography.

2.7 EDM in Higher Dimensions.

2.8 Rectilinear Histograms.

2.9 Depth and Multivariate Binning.

2.9.1 Data Depth.

2.9.2 Aside: Depth­Related Topics.

2.9.3 Annotated Bibliography.

2.10 Conclusion.

3 Partitions and Piecewise Models.

3.1 Divide and Conquer.

3.1.1 Why Do We Need Partitions?

3.1.2 Dividing Data.

3.1.3 Applications of Partition­based EDM Summaries.

3.2 Axis­Aligned Partitions and Data Cubes.

3.3 Nonlinear Partitions.

3.3.1 Annotated Bibliography.

3.4 DataSpheres (DS).

3.4.1 Layers.

3.4.2 Data Pyramids.

3.4.3 EDM Summaries.

3.4.4 Annotated Bibliography.

3.5 Set Comparison Using EDM Summaries.

3.5.1 Motivation.

3.5.2 Comparison Strategy.

3.5.3 Statistical Tests for Change.

3.5.4 Application – Two Case Studies.

3.5.5 Annotated Bibliography.

3.6 Discovering Complex Structure in Data with EDM Summaries.

3.6.1 Exploratory Model Fitting in Interactive Response Time.

3.6.2 Annotated Bibliography.

3.7 Piecewise Linear Regression.

3.7.1 An Application.

3.7.2 Regression Coefficients.

3.7.3 Improvement in Fit.

3.7.4 Annotated Bibliography.

3.8 One­Pass Classification.

3.8.1 Quantile­Based Prediction with Piecewise Models.

3.8.2 Simulation Study.

3.8.3 Annotated Bibliography.

3.9 Conclusion.

4 Data Quality.

4.1 Introduction.

4.2 The Meaning of Data Quality.

4.2.1 An Example.

4.2.2 Data Glitches.

4.2.3 Gaps in Time Series Records.

4.2.4 Conventional Definition.

4.2.5 Times Have Changed.

4.2.6 Annotated Bibliography.

4.3 Updating DQ Metrics: Data Quality Continuum.

4.3.1 Data Gathering.

4.3.2 Data Delivery.

4.3.3 Data Monitoring.

4.3.4 Data Storage.

4.3.5 Data Integration.

4.3.6 Data Retrieval.

4.3.7 Data Mining/Analysis.

4.3.8 Annotated Bibliography.

4.4 The Meaning of Data Quality Revisited.

4.4.1 Data Interpretation.

4.4.2 Data Suitability.

4.4.3 Dataset Type.

4.4.4 Attribute Type.

4.4.5 Application Type.

4.4.6 Data Quality – A Many Splendored Thing.

4.4.7 Annotated Bibliography.

4.5 Measuring Data Quality.

4.5.1 DQ Components and Their Measurement.

4.5.2 Combining DQ Metrics.

4.6 The DQ Process.

4.7 Conclusion.

4.7.1 Four Complementary Approaches.

4.7.2 Annotated Bibliography.

5 Data Quality: Techniques and Algorithms.

5.1 Introduction.

5.2 DQ Tools Based on Statistical Techniques.

5.2.1 Missing Values.

5.2.2 Incomplete Data.

5.2.3 Outliers.

5.2.4 Time Series Outliers: A Case Study.

5.2.5 Goodness­of­Fit.

5.2.6 Annotated Bibliography.

5.3 Database Techniques for DQ.

5.3.1 What is a Relational Database?

5.3.2 Why Are Data Dirty?

5.3.3 Extraction, Transformation, and Loading (ETL).

5.3.4 Approximate Matching.

5.3.5 Database Profiling.

5.3.6 Annotated Bibliography.

5.4 Metadata and Domain Expertise.

5.4.1 Lineage Tracing.

5.4.2 Annotated Bibliography.

5.5 Measuring Data Quality?

5.5.1 Inventory Building – A Case Study.

5.5.2 Learning and Recommendations.

5.6 Data Quality and Its Challenges.

O autorze

TAMRAPARNI DASU, PhD, and THEODORE JOHNSON, PhD, are both members of the technical staff at AT&T Labs–Research in Florham Park, New Jersey.

Blurb

A unique, integrated approach to exploratory data mining and data quality

Data analysts at information–intensive businesses are frequently asked to analyze new data sets that are often dirty–composed of numerous tables possessing unknown properties. Prior to analysis, this data must be cleaned and explored–often a long and arduous task. Ensuring data quality is a notoriously messy problem that can only be addressed by drawing on methods from many disciplines, including statistics, exploratory data mining, database management, and metadata coding.

Where other books on data mining and analysis focus primarily on the last stage of the analysis procedure, Exploratory Data Mining and Data Cleaning uses a uniquely integrated approach to data exploration and data cleaning to develop a suitable modeling strategy that will help analysts to more effectively determine and implement the final technique.

The authors, both seasoned data analysts at a major corporation, draw on their own professional experience to:

• Present a brief overview of the main analytical techniques used in data mining practices, such as univariate and multivariate summaries of attributes and their interactions including Q –Q plots, fractal dimension and histograms, nonparametric approaches incorporating data depth, and more
• Provide numerous references to the related literature on clustering, classification, regression, and more
• Focus on developing an evolving modeling strategy through an iterative data exploration loop and incorporation of domain knowledge
• Address methods of detecting, quantifying (metrics), and correcting data quality issues that significantly impact findings and decisions, using commercially available tools as well as new algorithmic approaches
• Use case studies to illustrate applications in real–life scenarios
• Highlight new approaches and methodologies, such as the DataSphere space partitioning and summary–based analysis techniques

A groundbreaking addition to the existing literature, Exploratory Data Mining and Data Cleaning serves as an important reference for data analysts who need to analyze large amounts of unfamiliar data, operations managers, and students in undergraduate or graduate–level courses dealing with data analysis and data mining.