Statistics Research

Research

Home People Research Awards Employment Contact

Our research covers a broad spectrum in theory, methodology and computing in statistics and applied probability, including a rich variety of collaborations with other research disciplines. As part of the Information and Software Systems Lab in AT&T LAbs, we have strong connections with computer scientists, mathemeticians, database experts, and leading researchers in software and programming languages. Our research cuts across the entire spectrum of statistical application and theory: time series, spatial statistics, Bayesian analysis, Markov models, machine learning, artificial intelligence and applied probability. Our research projects are typically motivated by the real-world issues generated in our dynamic industry. Current examples of active projects include:   Recommender Systems and Customer Preference Models. Recommender systems are models which predict items that a customer may like, based on previous purchase or rating behavior. Over the past few years we have built collaborative filtering models to predict movie and television preferences. This work resulted in winning two Progress Prizes as part of the $1 million Netflix Prize Competition. Interactive Data Analysis Visualization of data and the ability to interact with the data are powerful mechanisms for learning about data sets and doing exploratory statistical analysis. GGobi and iPlots are two such tools developed by in part by members of our department. We also work closely with our colleagues in the Information Visualization research group. Media Consumption Modelling. AT&T's data allows us insight into how consumers use different types of entertainment and media. We attempt to model this customer usage by accounting for the correlations in usage among TV, internet and cell phone usage.. Fraud Detection in Telecommunications. AT&T is the industry leader in protecting our networks from fraud. Fraud occurs in many forms: hackers breaking into lines, subscription fraud, spoofed cell phones, etc. Each of these requires large-scale data mining to determine the proper model fro detecting the fraud while minimizing false positives. Social Networks. Our communications network defines a rich social network in both the consumer and business spaces. Our research focuses on extracting information through these networks by looking at information flow through the network or viral marketing opportunities. For an example, see our work on Proximity Graphs and our paper on network based marketing. Time Series Clustering / Anomaly Detection in Data Streams. Monitoring vast data streams is a constabnt problem for large corporations. We model the incoming data as a multivariate data stream and build models to cluster and alert on those time series. Statistical Computing with R. Our department contains the researchers responsible for the S programming language. Our rich history in statistical computing continues to this day with development of statistical tools for the open source successor to S, known as R. Our specific focus is interactive data analysis tools, and interfaces with other languages . Data Quality As we gather and store increasingly complex, dynamic and massive data streams through automated means, our control and understanding of the data are getting strained. Managing, monitoring and cleaning the data prior to use by sophisticated algorithms, whether for network operations or marketing campaigns, is critical. Modelling data quality draws on expertise in Statstics, AI, Software engineering and Database Research to develop efficient analyses, algorithms and tools for data auditing, cleaning and repair. Applied Probability/Fluid Flow Models Fluid flow models are used in a wide variety of areas: network performance; storage and inventory processes; financial and insurance risk models. Recently, they have been used as models in the area of forest fire containment. An active area of research in the department is the development of powerful algorithms for stochastic fluid flow models to determine their time dependent performance measures, and their applications to the performance of high speed networks and call centers. We wrote the book on it! Recently, members of our department have written well-regarded books on their topics of interest. Take a look: Interactive Graphics for Data Analysis: Principles and Examples by Martin Theus and Simon Urbanek. Discusses exploratory data analysis (EDA) and how interactive graphical methods can help gain insights as well as generate new questions and hypotheses from datasets. Interactive and Dynamic Graphics for Data Analysis: With R and GGobi by Dianne Cook and Deborah Swayne. Describes the use of interactive and dynamic graphics as part of multidimensional data analysis. Exploratory Data Mining and Data Cleaning by Tamraparni Dasu and Theodore Johnson. Serves 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. An Introduction to Matrix Analytic Methods in Stochastic Modeling by V. Ramaswami and Guy Latouche. The authors present the basic mathematical ideas and algorithms of matrix analytic theory in a readable, up-to-date, and comprehensive manner. The New S Language: A Programming Environment for Data Analysis and Graphicsby Richard A. Becker, John M. Chambers, and Allan Wilks. Our history See our history page for information about a research tradition in statistics that goes back to Shewhart and Tukey.

Legal Notice © 2009 AT&T Intellectual Property. All Rights Reserved. AT&T and the AT&T logo are trademarks of AT&T Intellectual Property.