Close

We have launched a new CSE website. Please visit us here: http://computerscience.engineering.unt.edu

2014 UNT Graduate Exhibition

CSE Graduate Students participated in the UNT Graduate Exhibition on March 1, 2014. The Graduate Exhibition celebrates research in all its aspects as an essential and exciting part of graduate education at the University of North Texas. The Graduate Exhibition places special emphasis on communicating research and creative endeavor to a general audience and offers an unusual opportunity for professional development by challenging graduate students to present their work in clear, comprehensible terms to people outside their fields.

The Graduate Exhibition is also an opportunity for graduate students to see themselves as part of the larger University community, to share their creativity, and to appreciate the breadth of quality research being done at the University of North Texas.

The Department of Computer Science and Engineering had the following winners in the Computer Science and Information Technology category:

1st Karen Mazidi
2nd Mohamed Fazeen Mohamed Issadeen
3rd Sultanah Mohammed Alshammari


Sultanah Alshammari

Sultanah Mohammed Alshammari

Major Professor: Dr. Armin Mikler

Title: Data Mining Techniques for Predicating Breast Cancer Survivability among Women in the United States

Abstract: Breast cancer is the most common invasive cancer in females worldwide. According to the National Cancer Institute (NCI), it is estimated that in the U.S., 232,340 women will be diagnosed with breast cancer and 39,620 women will die from it in 2013. An accurate prediction of breast cancer survivability is very important in order to direct the treatment process and provide reliable indications for the public health authorities. In this study we used the breast cancer database provided by SEER program to investigate the use of data mining techniques, to automatically predict breast cancer survivability. The obtained results are promising for applying the data mining methods into the survivability prediction problem. These results suggest that among the predication algorithms tested, the C4.5 decision tree algorithm achieved the highest results with an accuracy of 88.96%. Moreover, a set of features that contributed the most in prediction accuracy was identified.


Garima Bajwa

Garima Bajwa

Major Professor: Dr. Ram Dantu

Title: Non-Invasive Monitoring of Cerebral Blood Flow using Brain Waves

Abstract: Cerebral blood flow is the measure of blood supply to the brain in a given time and typically varies between 50 to 54 ml/100g/min (100g of brain tissue). Brain needs continuous adequate supply of blood for its proper functioning. Cerebral Autoregulation (CA) refers to the intrinsic ability of the brain to maintain constant blood flow despite changes in systemic blood pressure. A major limitation in measuring the regulation of cerebral blood flow is the lack of a gold standard for its assessment. Our aim is to evaluate the possibility of using brain waves or Electroencephalograms (EEG) signals as one non-invasive methodology for the quantitative interpretation of CA. Also, the portability and ease of recording the EEG on a mobile platform makes it possible to monitor dynamic autoregulation continuously in healthy and diseased subjects.


Fahmida Hamid

Fahmida Hamid

Major Professor: Dr. Paul Tarau

Title: Apply Topic Modeling Techniques for Keyword Extraction and Word Sense Disambiguation

Abstract: Word Sense Disambiguation(WSD) and Keyword Extraction are two popular research fields in Natural Language Processing(NLP) whereas Topic Modeling is an emerging area in Machine Learning(ML). Recent developments on topic modeling have made it a popular tool for unsupervised analysis on large collections of literature. These models posit a set of latent topics, multinomial distribution over words, and assume that each document can be described as a mixture of these topics. We have chosen Latent Dirichlet Allocation(LDA) and its extensions: Correlated Topic Modeling(CTM), Hierarchical LDA(HLDA), Pachinko Allocation Modeling(PAM) as our tools to test the performance of automated topic modeling. After studying their differences, we propose to compare their performances on different domains of knowledge; keeping WordNET as a source of background, apply them for keyword extraction and word sense disambiguation on any set of documents. We believe blending these models with other techniques will be beneficial to NLP.


Sultanah Alshammari and Saratchandra Indrakanti
Saratchandra Indrakanti with Dr. Mary Harris

Saratchandra Indrakanti

Major Professor: Dr. Armin Mikler

Title: Computational Epidemiology Literature Classification Using Machine Learning Techniques

Abstract: Due to the multidisciplinary nature of Computational Epidemiology it can be challenging for researchers to identify and obtain relevant papers from the bulk of scientific literature. To alleviate this problem, we propose the use of machine learning techniques to automatically recognize computational epidemiology papers based on specific elements from the paper including; title, authors, keyword list, and the abstract. The aim of this work is to find effective classification techniques, the most informative sections of the article, the appropriate number of features, and the best feature representation method. The Random Forest classification algorithm achieved the highest results with an accuracy of 76%. Moreover, the dataset of authors names proved to be the best section for correctly classifying the papers, and using a binary feature representation worked better than utilizing a continuous feature representation. For future work, a full Computational epidemiology information retrieval and data extraction system will be developed.


Mohamed Fazeen Mohamed Issadeen

Mohamed Fazeen Mohamed Issadeen

Major Professor: Dr. Ram Dantu

Title: Another Free App: Does It Have the Right Intentions?

Abstract: Due to the increase in the use of mobile smart devices in recent days, security and privacy are important factors to be addressed. We propose a framework to identify malware in mobile applications by comparing the intention of an application with the permission requests.


Yiheng Liang

Yiheng Liang

Major Professor: Dr. Armin Mikler

Title: Graphical Models and Research of Causality on Epidemiological Data: A Computational Approach

Abstract: In almost all research areas of science on data, discovering causal relationship is important. Causal reasoning is not just empirical but analytical as well. Especially in epidemiology, studies of identifying causes and effects for diseases are imperative in that only through controlling real causes can diseases be prevented, and risks of diseases be mitigated. Moreover, due to the nature of data in epidemiology, most publicly available data are not experimental, but observational. Consequently new methods on large data should be developed for the discovery of causations in absence of temporal relationships.

Bayesian network can be a useful tool to study relations of factors and diseases, and their potential causal relationships. Through a computational approach with graphical representation of Bayesian network, this research provides new methodology and application of causal discovery on observational data with heuristics, from qualitative to quantitative reasoning.


Karen Mazidi

Karen Mazidi

Major Professor: Dr. Rodney Nielsen

Title: Automatically Generating Questions from Text

Abstract: As students read expository text, comprehension is improved by pausing to answer questions that reinforce the material. This project showcases a computer program that automatically generates questions from the text itself. First, expository text is read into a parsing program which determines the part of speech for each word, the sentences syntactic tree structures, as well as the semantic arguments of each predicate (e.g., agent, theme, instrument). Next, each parsed sentence is matched against templates to produce questions. For example in the sentence: Positive feedback increases the effect of a stimulus on the body, the noun phrase positive feedback is the agent, and the answer to the generated question: What increases the effect of a stimulus on the body? By carefully designing the templates, quality questions can be generated to assist students in understanding and remembering what they have read.


Venkata Kishore Neppalli

Venkata Kishore Neppalli

Major Professor: Dr. Cornelia Caragea

Title: Geo-mapped Sentiment Analysis on Twitter During Disaster Events

Abstract: Sentiment analysis has been widely researched in the domain of online review sites with the aim of getting summarized opinions of product users about different aspects of the products. However, there has been little work focusing on identifying the polarity of sentiments expressed by users during disaster events. Identifying sentiments expressed by members in an online social networking site can help understand the dynamics of the community, e.g., the main users’ concerns and the emotional impacts of interactions among members. Data produced through social networking sites is seen as ubiquitous, rapid and accessible, and it is believed to empower average citizens to become more situationally aware during disasters and coordinate to help themselves. In this work, we perform sentiment classification of user posts in Twitter during the Sandy hurricane and visualize these sentiments on a geographical map centered around the hurricane. We show how users’ sentiments change according not only to the locations of the users, but also based on the relative temporal distance from the disaster.


Shanti Thiyagaraja

Shanti Thiyagaraja

Major Professor: Dr. Ram Dantu

Title: Finger Blood Flow Monitoring Using Smart Phones

Abstract: The use of smart phones in healthcare applications is growing steadily. The inbuilt sensors are used to estimate the value of physiological data from human body. With progressive innovation, smart phone based medical applications will continue to be developed at an exponential rate. In this paper, we show that a smart phone can be used to monitor the blood flow in finger, based on the pulse height obtained from the fingertips. This is achieved by using the camera lens and the flash light of the smart phone. The height of the pulse rises along with the surrounding temperature indicating that the blood flow increases when the temperature becomes warmer. This study shows that there is a potential to monitor regulation of body temperature using smart phone.