Entries tagged with “thesis” from Computer Science Department News
Achrekar’s PhD research involved online social network (OSN) analysis, with a focus on predicting flu trends and information extraction from messages posted on OSN's. He was utilizing information posted on online social networks such as Twitter and Facebook to help improve the prediction of influenza levels within US population and was tracking of its spread.
He spend three years developing and perfecting an in-house framework called the Social Network-Enabled Flu Trends (SNEFT). The software that uses continuous data-collection mechanism to monitor flu-related messages, extract relevant user demographic and location information, classify messages in real time, and predict the current influenza levels.
Starting 2009, Achrekar tapped into Twitter and Facebook and extracted tens of millions of influenza-related user posts till date, to provide an almost-instantaneous snapshot of current epidemic conditions and building comprehensive mathematical models that improves the estimate of nationwide flu activity.
Achrekar's approach can significantly enhance public health preparedness against influenza and other large-scale pandemics.
For this research, Achrekar was supported with a $200,000 grant from the National Institutes of Health under a Small Business Innovation Research Award.
His results have been presented in various scientific publications:
- His paper “Twitter improves Seasonal Influenza Prediction" at the fifth annual International Conference on Health Informatics, Portugal, February 2012, received the Best Student Paper Award. This is a prestigious and competitive conference with a 9% full-paper acceptance ratio.
- His paper titled “Predicting Flu Trends using Twitter Data” at the International Workshop on Cyber-Physical Networking Systems (CPNS) 2011, in conjunction with IEEE INFOCOM 2011, held in Shanghai, China, April 10–15, 2011, is considered to be a foundational research for flu tracking with OSN data and has received many citations.
- The paper titled “A Spatio-Temporal Approach to the Discovery of Online Social Trends” at the fifth Annual International Conference on Combinatorial Optimization and Applications (COCOA), Zhangjiajie, China in August 2011 provides insights into using online social network data to discover trends in other domains.
- Also “Vision: Towards Real Time Epidemic Vigilance through Online Social Networks” was his first ACM Workshop paper at Mobile Cloud Computing & Services: Social Networks and Beyond (MCS), in conjunction with ACM MobiSys, San Francisco, California, June 2010, that explains the SNEFT architecture for flu tracking and pandemic outburst detection using online social network data.
Dr. Achrekar’s committee included Cindy Chen (Computer Science), Georges Grinstein (Computer Science), Yan Luo (Electrical and Computer Engineering), and Jie Wang (Computer Science).
Harshavardhan (Harsh) Achrekar at his doctoral defense on September 6, 2012.
Penta’s work was inspired by an experience he had teaching video game design in a summer camp held at the university. Two of his students were trying to position a cannon ball at the end of a cannon, which could be positioned at various angles. Near the end of a day, the students asked Penta how to do this, and he told them, “That is just a bit of trigonometry. I will show you tomorrow.”
Penta was then surprised when, the following day, the students had solved the problem on their own. As he described it:
I learned that they had gone home and introduced themselves to trigonometry by searching the web. ... These students had taken responsibility for their learning, and became self-directed problem solvers. They had taken a subject disliked by most students and [...] learned the essence of an important math concept. ... They were motivated by their own problem, a problem within a context about which they cared, the game they were making.Penta used this insight as a jumping-off point for his Master’s project. He set out to develop an intentional learning environment where students would be encouraged to build their mathematical competencies through video game creation.
He then evaluated three different learning environments: an in-school mathematics classroom, an after-school game design workshop, and an after-school mathematics-focused game design workshop. Using a design-based research methodology, Penta created a series of evaluation tools to measure students’ learning, and refine the learning environment in each iteration.
Ultimately, Penta argued that because of curricular constraints, in-school time is not suitable for student video game design projects. He concluded that interventions should be structured around authentic video game design with integrated, focused mathematical design challenges. Finally, he demonstrated that students developed improvements in their understanding of mathematical concepts including plotting Cartesian coordinates, using negative numbers, and finding functions from patterns.
Penta’s work was advised by Prof. Fred Martin. Douglas Prime (College of Engineering) and Prof. Marvin Stick (Mathematical Sciences) were thesis readers. A copy of the thesis is available on Proquest or as a local PDF.
“Haunted Mansion,” a student-created game in response to the maze challenge. Student games had to have a “hero” character which moved using the arrow keys, and at least two “good” and two “bad” non-player characters (NPCs). When the hero struck a good NPC, its number of lives had to increase, and when it collided with a bad NPC, it would lose a life. When all lives were lost, the game had to end.
Lipman’s research focused on addressing the challenge of developing client-server web systems.
The skill sets required for these two pieces are different. Often, the front-end and back-end are developed and tested completely independently, based purely on an interface specification.
Lipman addressed this by developing his framework, LIBERATED, which stands for “Lipman’s In-Browser EnviRonment for Application TEsting and Development.”
In the thesis, Lipman proposed a new methodology for web-based client-server application development, in which a simulated server is built into the browser environment to run the back-end code.
This design allowed the front-end code to issue requests to the back-end in either a synchronous or asynchronous fashion, and single-step, using a debugger, directly from front-end code into back-end code, thereby completely testing both components with the desktop browser environment.
In Lipman’s system, that exact same back-end code, now fully tested in the simulated environment, is then recompiled and moved to a real server.
In the defense, Lipman presented the detailed design of LIBERATED, and described how he used it to develop the App Inventor Community Gallery, a web system created for users of Google’s App Inventor programming environment for Android phones to share their projects.
Prof. Fred Martin served as Lipman’s thesis adviser, and Dr. Mark Sheldon served as his thesis reader. Lipman’s research was supported by a grant from Google.
Baumann’s research focused on the design and development of Weave, a web-based data visualization framework that is now available under an open source license. Dr. Baumann oversaw the development of this software package from its original design to the current implementation with his research on a novel windowing environment for web-based data visualization that allows transitions between many types of user interactions and layouts.
Baumann’s work was funded by the Open Indicators Consortium (OIC), which was founded to both develop this platform and offer a community of learning for not-for-profits and government agencies who deal with indicator data, or custom measures that track progress towards a goal or quality of an entity. The agile development process was used to provide the members with regular releases and use their feedback to drive feature design and evolution.
Dr. Baumann extended many of the concepts of the earlier desktop-based Universal Visualization Platform in whose development he participated. He led the team in its first Weave designs, and through the feedback from the OIC, extended that design to provide a more flexible and customizable framework for web-based data visualizations. The new design supports dynamic customizable layouts, visualizations targeted to different levels of users, and exploratory data visualization, all within a novel windowing environment.
Prof. Grinstein noted that there are already many users of the software ranging from small communities to cities such as Boston, Seattle, Chicago and Atlanta, to states such as Rhode Island, Massachusetts and Connecticut, with many more anticipated users.
Dr. Baumann’s dissertation committee readers were Dr. William Mass (Economic and Social Development of Regions) and Dr. Haim Levkowitz (Computer Science). Baumann's thesis document is archived on ProQuest.
Agnello developed an Android application that tracks a user’s physical activity using sensors built into most Android phones (primarily, GPS and accelerometer data). The application then analyzes the collected data and verifies that the user is actually doing physical activity (e.g., identifying running vs. sitting in a moving car).
Upon ending a session (shutting down the application or user selection), the application communicates with a web server and sends the user’s current physical activity completed. A corresponding web site displays different measurements of the user’s physical activity (by week, month, or year). Additionally, the web site offers RSS feeds dedicated to help motivate the user to continue their physical activity.
As this is application part of a framework for what could be a complete solution, Agnello designed this project in a modular structure for easy porting. The technologies used in the development of this work were the Android operating system, PHP, MySQL and HTML. Dr. Guanling Chen served as thesis reader for the project.
Fertitta’s work involved the development of a set of custom apps that were used in a high school physics classroom. Fertitta worked closely with a local high school teacher to conceive of the apps, and then implemented them and supported the teacher in using them in this classroom.
Fertitta's project extended the Engaging Computing Group's Internet System for Networked Science Experimentation (iSENSE), which Fertitta also helped develop. In his thesis work, Fertitta's apps allowed students to gather acceleration data on Android smartphones. These data were then uploaded to iSENSE, and then students collaboratively made sense of the data by overlaying views of their various data sets.
In one of the projects, students went on rides at the Canobie Lake amusement park, and used Fertitta's app to collect acceleration data. Then, back in the classroom, the students completed worksheets where they predicted what their graphs would look like. Finally, students viewed the actual data, and had to figure out which graph matched which ride.
In analyzing student work and in a post-interview with the teacher, Fertitta argued that in this case “smartphones were ‘far superior’ to other technologies” for data collection, and that the use of the iSENSE system, which easily allowed students to overlay each other’s data, “facilitated more in-depth discussion” than other tools.
Fertitta's thesis was advised by Profs. Fred Martin (Computer Science) and Michelle Scribner-MacLean (Graduate School of Education). Fertitta's work was supported by grants from the National Science Foundation (DRL-0546513 and DRL-0735597) and a gift from Google Inc. A copy of the thesis is online here.
Student prediction graph (left) and actual data from the iSENSE web system. Also, interact with the visualization live on the iSENSE site at this URL: http://isense.cs.uml.edu/visdir.php?id=121
The use of unbounded point sets requires special handling in computational geometry. One method is to calculate or estimate a maximum volume of space in which all geometric manipulations will fit: a bounding box. Alternatively there is a technique called infimaximal frames. Neither option was suitable for use in this case in the field’s standard Computational Geometry Algorithms Library, and so M’Sadoques’ algorithm was developed as a solution.
In addition to two alternate proofs, additional work was done to allow the packing of two objects into a container, again avoiding unbounded sets. The algorithms were implemented in C++ and timings were taken using simple 3D input shapes.
Present at the defense were thesis committee members Drs. Karen Daniels, George Grinstein and Cindy Chen from the Computer Science department, and remotely Dr. Victor Milenkovic from the University of Miami.
The complete thesis may be retrieved from the department’s Technical Report repository. The results of this work will also be presented in the Cutting and Packing stream of the Conference for the International Federation of Operational Research Societies (IFORS 2011) in Melbourne, Australia on July 11.
Image illustrating containment problem. Beyond two shapes, the containment problem is NP-complete.