Data Science Lab
Director:
Nathalie Japkowicz
The
Computer Science Data Science Lab at American University develops and applies
machine learning techniques to a variety of domains. Current areas of interest
include: data stream mining, anomaly detection, concept drift, class
imbalances, one-class learning. Current domains of application include:
cyber-security, mobile traffic classification, and genomics.
Current Members
Senior Personnel
Roberto Corizzo, Assistant
Professor of Research from July 2019 until June 2020
Roberto Corizzo is an assistant professor of research in the Department of Computer Science at
American University, where he is pursuing research in lifelong streaming anomaly detection.
Prior to that, he was a research fellow in the Department of Computer Science at University of
Bari, Italy, and a research intern at the INESC TEC research institute in Porto, Portugal, under
the supervision of Prof. Joao Gama. He co-authored 16 articles, including 5 journal publications
in venues such as IEEE Transactions on Industrial Informatics, Data Mining and Knowledge
Discovery, and Information Sciences.
Main Project: Hierarchically and Laterally Growing
Autoencoder (HLGAE)
(Funded by DARPA)
Lifelong learning allows to create adaptive systems that evolve over time, adapting to changing
environmental conditions. This behavior is possible by combining neural networks and machine
learning techniques with statistical learning theory. Our aim is to extract adaptive models with
growing capabilities in two modalities: lateral and hierarchical growth. Most data streams
contain sub-concepts, and concept or subconcept drifts occur naturally as time evolves. The
hypothesis is that a growing hierarchy of models can improve the recognition of diverse sub
concepts, and, thus, enable a more precise anomaly detection performance according to the most
specialized and fitting sub-model.
Students
Reham Amin,
Visiting PhD Student: from July 2019 until end of December 2019
Reham Amin is a visiting PhD student from the Faculty of Computers and Informatics, Suez
Canal University, Ismailia, Egypt. Her research combines elements of cybersecurity, machine
learning and information visualization. At AU, she is crating a visualization system for
hyperparameter tuning in neural networks.
Yohan Dauphin, Visiting Scholar: from September 2019
until end of February 2020
Yohan Dauphin is a visiting scholar majoring in Computer Science from the engineering school
of CPE Lyon in France. His research focuses on machine learning and deep learning for
detection of abnormalities in medical images.
Victor H. Barella, visiting
Ph.D. student (University of São Paulo, Brazil): September 2019 - May 2020
Victor Barella is a Ph.D. student in the Institute of
Mathematical and Computer Sciences at the
University of São Paulo (ICMC-USP), Brazil, where Dr. Andre de Carvalho supervises him. His
current research focuses on data characteristics and pre-processing techniques for imbalanced
classification tasks. His research interests include imbalanced datasets, data complexity
measures, meta-features, pre-processing techniques, and hierarchical classification. At American
University, he is working on a meta-learning approach for imbalanced classification tasks under
the supervision of Dr. Nathalie Japkowicz.
Alexis
Godwin, American University, BS (expected 2020)
Alexis Godwin is a current senior at American University majoring in Computer
Science. She has interests in data mining, remote sensing and geospatial data, and
data analytics. She is currently working with statistical and machine learning methods
on medical data.
Past Members
Senior Personnel
Zhen
Liu, Visiting Scholar from March 2018 to March 2019
Zhen
Liu is a visiting scholar from the School of Medical Information Engineering at
Guangdong Pharmaceutical University
in Guangzhou, China where she has been working
as a Lecturer. Her research
interests lie in the areas of Machine Learning and Cyber
Security. Her previous research was
in the areas of multi-class learning, the class
imbalance problem applied to mobile
traffic classification. At American University, she
is focusing on Anomaly Detection in
Data Streams, and devising methods for dealing
with issues of concept drifts in the
context of intrusion detection.
Li Liu,
Visiting Scholar from September 2018 to February 2019
Li Liu is a visiting scholar from the School of Information Science and
Technology
at Huizhou University in Guanzhou, China where
she has been working as an Associate
Professor. Her research interests lie in the areas of Machine Learning
and Data Mining.
Her previous research was
in the area of dimensionality reduction, mainly focusing on
manifold algorithms, and applied to
content-based image retrieval. At American University,
she is focusing on one-class classification and devising methods for
dealing with issues
of sub-concept drifts in complex domains.
Zhao Yang, Postdoctoral fellow from
January 2017 until July 2018
Zhao Yang conducts
research mainly in big data analytics, high performance computing
and
statistical learning theory. Among his non-professional interests are hiking,
swimming, etc.
He is
the winner of 2015 Alan Berman Research Publication Award, Naval Research
Laboratory (NRL),
Washington
D.C.(Best Paper Award of the Department of Navy)
Projects:
High
Performance Data Mining Analytical Environment for Large Scale Cyber Security
Data
In this project, we propose a
framework for processing and analyzing large-scale cyber security data
using a Big Data infrastructure.
Existing Big Data solutions do not include high performance
mechanisms to analyze large-scale
cyber-security data. In this work, we extend current open-source
platforms to support cyber-security
data and demonstrate its analytical use with some common data
types and data mining technology
provided by the open source solution. The resulting framework is a
robust capability to share large-scale
security data and make its outputs available to end users.
Efficient 3-D
Object Detection for Large Scale DEM (Digital Elevation Model) data set
We propose a three-dimensional object detection method using
large scale DEM data.
Recently, 2D object detection has been widely used in
navigation and geospatial
computation. Nevertheless, conventional object
detection systems such as edge detection
algorithms was originally designed
as an image processing technique for finding the
boundaries of objects within images. The edge
detection algorithms have low
universality because they are designed for 2-D objects,
which limits the types of targets
that can be detected. Our method implements machine learning
algorithms that solves
this problem and enables high-efficient, deterministic
measurement of the feature of a
large scale 3D target from the DEM data
set without any prior knowledge about the
feature of the targets. Using this technique and a
prototype system that we developed, we
also demonstrated a number of
applications, including sea mountain detection which can
be used by surface vessels and UUV (unmanned
underwater vehicles),etc.
Students
Jonathan Kaufmann is a senior in the
department of Computer Science. His research
focuses on natural language
processing and genomics. His previous research examined
the economic impact of war on child
soldiers, as well as reintegration.
Roberto Corizzo, visiting Ph.D. student
(University of Bari, Italy): Spring-Summer 2017
Roberto
Corizzo is currently finishing his Ph.D. in predictive
models for streams of sensor data in the
Department
of Computer Science at the University of Bari, Italy. He is part of the
KDDE research group
coordinated
by Prof. Donato Malerba. His research interests
include big data analytics, data mining and
predictive
modeling techniques for sensor networks.
