Thesis Proposal Presentations
Sipan Ye
M.Sc. Thesis Proposal
Supervisor: Dr. Yuanzhu Chen
Inducing social interaction through co-location using radio beacons
Department of Computer Science
Thursday, February 23, 2017, 2:00 p.m., Room EN 2022
Abstract
The ability in analyzing and predicting human social structures and behaviors is essential in sociology and anthropology. Originally, scientists have conducted research by manually observing people’s daily activities, movements, and communication [27]. This is laborious and error-prone. More and more scientists generate their researches through these approaches, like GPS-based data model [8] [17]. However, these methods are not a perfect representation of people’s behaviours and activities on stability and convenience, such as people they met, locations they visited, and the amount of time they stayed in a place. These data are important for activities of humans in reality [9]. In our research, we want to improve the data performance and provide a more efficient approach for inducing social interaction. We plan to invite volunteers fitted with Bluetooth LE (Low Energy) transmitters. In addition, we will also allocate some transmitters in fixed locations to help us gather more specific location data. We propose a contact network model to provide social interaction of people based on these data.
Keywords: Bluetooth, iBeacon, Social Network, Contact Network.
Ali M.S. Alfosool
M.Sc. Thesis Proposal
Supervisor: Dr. Yuanzhu Chen
Geographical Mobile Social Network: Local Community Integration and Analysis
Department of Computer Science
Thursday, February 23, 2017, 2:20 p.m., Room EN 2022
Abstract
Non-profit organizations are vital to having a better life within a community we live in today. However, many communities have long suffered the absence of quality in the services (if at all) provided. This may partially be due to lack of allocated funding and the recent budget cuts; but it is also an indirect result of today’s lifestyles. Social networks and smartphones are now, more than ever, important parts of most people’s lives. We aim to combine the advantages of the two and integrate a geographical and mobile based social network within a local community and to analyze the network’s pattern. Although, there has been some previous work on geosocial networking (such as Brightkite), they have mainly focused through a business perspective to increase potential profit. Additionally, most of these businesses aim to attract members from all around the world and do not generate a fully local-based network. We are interested to study how people, in a local community-based social network, are connected and interact with one another with respect to time and their location. Our analysis will evaluate the network’s evolution over time which helps to determine specific characteristics of the locally-formed network. Additionally, Closeness, Betweenness and PageRank centralities of individuals can be measured to analyze members’ communication and friendship network as well as most in-fluential members of the community. Our work is based on collaborating with a local non-profit organization (East Coast Trail Association) and involves developing a mobile application that takes advantage of an array of locally developed services. Such as outdoor fitness activity tracking, offline trail mapping, and local geo-based social networking. This application addresses many free services that the community has longed-for, yet attracting more members and support for the local non-profit organization. In addition to safety, health, and geo-based social connection services provided by this app, the data collected (with users’ permission) from these services such as social, physical and geographical patterns, will be most beneficial to this research for further community pattern and statistical analysis.
Faramarz Dorani
M.Sc. Thesis Proposal
Supervisor: Dr. Ting Hu
Detecting gene-gene interactions in Colorectal cancer using data mining and machine learning approaches
Department of Computer Science
Thursday, February 23, 2017, 2:40 p.m., Room EN 2022
Abstract
The fundamental task of genome-wide association studies is to detect genetic variations that are mainly contributing to the disease state. To achieve this purpose, many research areas have been explored and machine learning methods have gained the most interests. On the other hand, colorectal cancer is a common cause of cancer deaths in developed countries and especially it has a high incidence rate in the newfoundland and Labrador province. Therefore, finding the affecting genetic factors can help better understand the disease in order to better treat and prevent the disease. In this thesis, we will investigate a few powerful machine learning algorithms on the colorectal cancer genetic data which have been collected from subjects in Newfoundland and Labrador. Random forests (RF) and logistic regression will be applied to the data to detect the underlying gene-gene interactions and the correlation of genetic variants with the disease. To overcome the challenge of high dimensionality in this big genomic dataset, we will investigate several feature selection methods to reduce the number of features in the dataset. Afterwards, the results of our analysis will be biologically validated to extract new knowledge in the understanding of the disease in order to better diagnose, treat and predict the disease.
Keywords: machine learning, gene-gene interactions, colorectal cancer, feature selection.