Web and Complex Systems Lab News Archive
Dept. of CSE, Wright State University
1/20/18: Lakshika successfully defends her thesis on studying gang member profiles on Twitter. She will be joining Gracenote as a research engineer in Okland, CA this summer.
1/16/18: Some (non-)Universal Features of Web Robot Traffic accepted at IEEE Conference on Information Sciences and Systems. This study identifies statistical properties of web robot traffic that are common and contrasting across multiple web servers around the world. Great job Mahdieh!
1/14/18: Contrasting Web Robot and Human Behaviors with Network Models accepted for publication in the Journal of Communications. The article contrasts web robot and human visititation behaviors through the lens of network analytics. Congratulations Kyle!
11/15/17: Relating Input Concepts to Convolutional Neural Network Decisions accepted at NIPS Workshop on Explainable Machine Learning. This work develops a novel algorithm to find distributed representations of input concepts in a convolutional neural network. It then studies how types of representations could affect the decisions the deep learning system makes. Well done Ning!
11/3/17: EmojiNet, our machine readable sense inventory for Emoji co-developed by Lakshika, has been selected for publication as a Kaggle dataset. Check it out here. If you play around with it let us know what you find! And, an article on EmojiNet has been picked up by Psychology Today. Neat!
10/10/17:Realistic Traffic Generation for Web Robots accepted at IEEE Intl. Conference on Machine Learning and Applications 2017. This work presents a generative process for producing synthetic sequences of web robot traffic. A stochastic process moderates temporal properties while a Bayesian model assigns behavioral attributes. The synthetic traffic impacts the performance of web server caches in similar ways as real robot traffic. Well done Kyle!
7/3/17:Explaining Trained Neural Networks with Semantic Web Technologies: First Steps accepted at the International Workshop on Neural-Symbolic Learning and Reasoning 2017. This paper documents our first (successful) efforts and attaching semantically driven explanations about why a deep network reaches a decision in scene classification tasks.
6/19/17:Seasonality in Dynamic Stochastic Blockmodels
accepted at the
Workshop on Complex Methods for Data and Web Mining at IEEE/ACM Web Intelligence 2017.
This work introduces a new staitsical network model where edge formation probabilities
depend on the `type' of each node and on seasonal time series processes that a latent
in data. An inference procedure to recover the latent seasonal processes from data is
also validated. This is Jace's first publication with WaCS!
6/5/17: A Soft Computing Approach for Benign and Malicious Web Robot Detection accepted in the journal Expert Systems with Applications. The article
describes a new approach to detect Web robot sessions from web logs. By
fuzzy set theory and Markov clustering, the approach selects an appropriate set of classification features, thereby adapting to server-specific session level patterns
automatically. This is Mahdieh's first publication with WaCS!
5/5/17: Matt Piekenbrock's proposal to the 2017 Google Summer of Code was accepted!
Matt will be funded by Google to implement an open-source
R package that unifies and improves algorithms for estimating the
empirical cluster tree of a dataset. The package will be a
realization of recent, rapid advances in density-based clustering.
The proposal selection process is very competitive. Congratulations Matt!
4/1/17: Dr. Doran's brief
monograph Network Role Mining
and Analysis has been publised under Springer's
Briefs in Complexity series. The book gives an overview
of historical, classic, and modern methods for the node
role discovery problem in networks.
3/7/17: EmojiNet: An Open Service and API for Emoji Sense Discovery,
accepted at the AAAI Intl. Conference on Weblogs and
Social Media. This work describes the release of our public knowledge base for interpreting the
sense, or meaningful interpretation, of emoji when used in a particular context. Applications to
emoji sense disambiguation and similarity evluation is discussed. Check out the
resource here -- well done Sanjaya and Lakshika!
2/1/17: A Soft Computing Prefetcher to Mitigate Cache
Degradation by Web Robots, accepted at the
2017 Intl. Symposium on Neural Networks. This work introduces our new prefetcher for web
resources requested by robots or crawlers, which combines a deep recurrent neural network
with Bayesian networks that combine prior global information with session-specific information about a robot.
Ning and Kyle are joint first authors on this work!
1/17/17: Dr. Doran's book chapter on
Graph/Link Mining has been
accepted for inclusion in Springer's upcoming Encyclopedia of Big Data. The
chapter covers an introduction to graph mining for a non-technical expert, relates
graph mining techniques to the scientific field of network science, and presents
other fundamental concepts and graph mining techniques.
12/20/16: Ning Xie wins two scholarships to attend important
events for women in computing: the
CRA-W Grad Cohort Workshop in Washington, DC and the
Ohio Celebration of Women in Computing at Lake Huron in Ohio. Congratulations Ning!
12/15/16: Nathan Rude successfully defends his MS Thesis
Intelligent Caching to Mitigate the Impact of Web Robots on Web Servers. He will be
taking a job as a Software Engineer for data intensive computing at LexisNexis Special Services this January.
12/13/16: Samir Yelne successfully defends his MS Thesis
Measures of User Interactions, Conversations, and Attacks in a Crowdsourced Platform Offering Emotional Support. He will be taking a job as a Data Scientist at Cisco in
San Jose, CA this January.
10/20/16: Keep the Conversation Going: Engagement-Based Customer Segmentation on Online Social Service Platforms, accepted in the
journal Information Systems Frontiers. This work integrates kernel functions
into the traditional k-means clustering algorithm to segment customers on online
platforms having social functions (heavy-tails abound). Nripesh is first author
on this work and represents his work during his Summer 2015 WaCS visit!
10/1/16: WaCS is awarded a research award from the Ohio Federal
Research Network titled Human Centered Big Data. This project
is in collaboration with the DaSE and BiRG labs at Wright State, as well as the Wright State
Research Institute, Case Western University, and the Ohio State University.
9/28/16: RFID-Based Information Visibility for Hospital Operations: Exploring its Positive Effects using Discrete Event Simulation,
accepted in the journal Health Care Management Science.
This article presents a simulation-based performance analysis of hospitals
wherein patients visit stations that have visible waiting times. Nathan
played a major role in this collaborative work
between WaCS and the Dept. of Information Systems and Supply Chain Management.
8/20/16: EmojiNet: Building a Machine Readable Sense Inventory for Emoji,
accepted at SocInfo 2016. This paper introduces a new web resource: a machine
readable sense inventory for emoji. EmojiNet integrates multiple emoji lexicographic resources found on the Web along with BabelNet, a comprehensive machine readable sense inventory for words, to infer sense definitions. Lakshika is second
author on this work and was instrumental in its development!
Check out the resource here!
8/15/16: Ning Xie wins an
NSF Travel Fellowship to attend the 12th annual
Reasoning Web Summer School in Aberdeen, Scotland.
The school will develop her knowledge in data semantics
and linked data theory and systems. The school may prove
useful in future explorations at the intersection of
deep learning and data semantics, and in web traffic and linked data analysis.
7/29/16: WaCS welcomes Jace Robinson (MS) and
Ethan Wolfer (UGrad) to our research group.
7/14/16: Finding Street Gang Members on Twitter,
accepted at IEEE/ACM ASONAM 2016.
This paper proposes a system to automatically identify
twitter profiles affiliated with street gangs. Its novelty lies in the use of
heterogeneous features, including image tags inferred by a deep neural
network, tags from YouTube video links, and emoji use, whch were inferred
by an analysis of what may be the largest set of twitter profiles related
to gang members. Lakshika is first author on this work!
6/22/16: Word Embeddings to Enhance Twitter Gang Member Profile Identification , accepted at IJCAI Workshop on Semantic Machine Learning.
This paper discusses the use of deep learning to map text, extracted from a
set of heterogeneous features, into a single space for Twitter profile classification. This is Lakshika's first
4/19/16: Measuring the Users and Conversations of a Vibrant Online Emotional Support System , accepted at IEEE Intl. Symposium on Computers and Communications. This paper follows up our ASONAM 2015 work on studying the users of a large-scale emotional support service and the dynamics of the conversations they hold with each other. Students Samir Yelne and
Nripesh Trivedi are co-authors on this work.
4/14/16: Matt Piekenbrock is listed as a co-author of the R package dbscan. The package implements a critical non-parametric clustering algorithm and is downloaded at least 2,300 times per month around the world. Download it from CRAN!
2/3/16: Exploring Information-Optimal Network Discretization for Dynamic
Network Analysis, accepted at Sunbelt 2016 . This poster
discusses how a notion of entropy defined by the structure of temporal networks may be
used to guide the discretization of continuous network data.
Sunbelt is the flagship conference for the International Network for Social Network Analysis.
12/14/15: A Runner-up Best Teaching
Paper Award is given to Teaching the Foundations of Data Science: An Interdisciplinary Approach at the 2015
SIGDSA Business Analytics Congress!
12/14/15: WaCS welcomes Ning Xie (PhD) and Scott Duberstein (UGrad) to
our research group.
12/11/15: Operationalizing Central
Place and Central Flow Theory With Mobile Phone Data, accepted
in the journal Annals of Data Science. This work
demonstrates how artifacts explained by Central Place and Central
Flow Theory, which are geographic explanations about how regions
develop economically and socially, may be unearthed in mobile
11/16/15: WaCS is awarded an REU Supplement to
our NSF project. The funds will support Logan Rickert and a new UGRA to
be hired through 2017.
10/15/15: Teaching the Foundations of Data Science: An Interdisciplinary Approach,
accepted at 2015
SIGDSA Business Analytics Congress! .
This paper documents the design, student
experiences, and outcomes of a progressive undergraduate course on data analytics for both CS and MIS students
co-taught by Dr. Doran. The Congress is a pre-event of ICIS, the largest
professional association for information systems.
9/18/15: Request Type Prediction for Web Robot and Internet of Things Traffic ,
accepted at IEEE ICMLA 2015. Towards building
predictive caches for web servers and clouds that can service robot and IoT traffic with better performance,
this work motivates the use of
recurrent neural networks to anticipate the type of a resource to be requested by bots or IoT devices.
This is Nathan's first publication with WaCS!
9/18/15: WAMINet: An Open Source Library for Dynamic Geospace Analysis Using WAMI , accepted at IEEE ISM 2015. We introduce an open source tool for processing WAMI imagery, and its method to derive network models of the dynamics of a monitored geospace. This is Matt M.'s first publication!
A best paper award at IEEE/ACM ASONAM 2015 awarded to
Stay Awhile and Listen: User Interactions in a Crowdsourced Platform Offering Emotional Support!
8/27/15: On the Discovery of Social Roles in Large Scale Social Systems, accepted in the journal Social Network Analysis and Mining. This theoretical work describes an unsupervised, data-driven method to extract the roles of actors in large scale social systems -- an advancement over prior art that defined qualitative analysis or used unscalable mathematical definitions.
7/10/15: Social Media Powered Human Sensing for Smart Cities, accepted in the journal AI Communications. This is an extension of Dr. Doran's 2013 work on social media mining in support of public utilities. This research was carried out in collaboration with industry partners.
6/22/15: Stay Awhile and Listen: User Interactions in a
Crowdsourced Platform Offering Emotional Support, accepted at IEEE/ACM ASONAM 2015. This may be the first study that quantitatively measures the behaviors and interactions of people on a large (>100k users) social system designed to offer emotional support to those in need. This is Samir's first publication!
5/12/15: WaCS welcomes Nripesh Trivedi, a visiting student from IIT Varanasi to our research group.