Digital Commons Network^™

A Classification Framework For Imbalanced Data, Piyaphol Phoungphol

Computer Science Dissertations

As information technology advances, the demands for developing a reliable and highly accurate predictive model from many domains are increasing. Traditional classification algorithms can be limited in their performance on highly imbalanced data sets. In this dissertation, we study two common problems when training data is imbalanced, and propose effective algorithms to solve them.

Firstly, we investigate the problem in building a multi-class classification model from imbalanced class distribution. We develop an effective technique to improve the performance of the model by formulating the problem as a multi-class SVM with an objective to maximize G-mean value. A ramp loss function …

Real-Time Physics Based Simulation For 3d Computer Graphics, Xiao Chen

Computer Science Dissertations

Restoration of realistic animation is a critical part in the area of computer graphics. The goal of this sort of simulation is to imitate the behavior of the transformation in real life to the greatest extent. Physics-based simulation provides a solid background and proficient theories that can be applied in the simulation. In this dissertation, I will present real-time simulations which are physics-based in the area of terrain deformation and ship oscillations.

When ground vehicles navigate on soft terrains such as sand, snow and mud, they often leave distinctive tracks. The realistic simulation of such vehicle-terrain interaction is important for …

Coronary Artery Calcium Quantification In Contrast-Enhanced Computed Tomography Angiography, Abinashi Dhungel

Computer Science Dissertations

Coronary arteries are the blood vessels supplying oxygen-rich blood to the heart muscles. Coronary artery calcium (CAC), which is the total amount of calcium deposited in these arteries, indicates the presence or the future risk of coronary artery diseases. Quantification of CAC is done by using computed tomography (CT) scan which uses attenuation of x-ray by different tissues in the body to generate three-dimensional images. Calcium can be easily spotted in the CT images because of its higher opacity to x-ray compared to that of the surrounding tissue. However, the arteries cannot be identified easily in the CT images. Therefore, …

A Framework For Discovery And Diagnosis Of Behavioral Transitions In Event-Streams, Arash Akhlaghi

Computer Science Dissertations

Date stream mining techniques can be used in tracking user behaviors as they attempt to achieve their goals. Quality metrics over stream-mined models identify potential changes in user goal attainment. When the quality of some data mined models varies significantly from nearby models—as defined by quality metrics—then the user’s behavior is automatically flagged as a potentially significant behavioral change. Decision tree, sequence pattern and Hidden Markov modeling being used in this study. These three types of modeling can expose different aspect of user’s behavior. In case of decision tree modeling, the specific changes in user behavior can automatically characterized by …

Viral Quasispecies Reconstruction Using Next Generation Sequencing Reads, Bassam A. Tork

Computer Science Dissertations

The genomic diversity of viral quasispecies is a subject of great interest, especially for chronic infections. Characterization of viral diversity can be addressed by high-throughput sequencing technology (454 Life Sciences, Illumina, SOLiD, Ion Torrent, etc.). Standard assembly software was originally designed for single genome assembly and cannot be used to assemble and estimate the frequency of closely related quasispecies sequences.

This work focuses on parsimonious and maximum likelihood models for assembling viral quasispecies and estimating their frequencies from 454 sequencing data. Our methods have been applied to several RNA viruses (HCV, IBV) as well as DNA viruses (HBV), genotyped using …

Data Collection And Capacity Analysis In Large-Scale Wireless Sensor Networks, Shouling Ji

Computer Science Dissertations

In this dissertation, we study data collection and its achievable network capacity in Wireless Sensor Networks (WSNs). Firstly, we investigate the data collection issue in dual-radio multi-channel WSNs under the protocol interference model. We propose a multi-path scheduling algorithm for snapshot data collection, which has a tighter capacity bound than the existing best result, and a novel continuous data collection algorithm with comprehensive capacity analysis. Secondly, considering most existing works for the capacity issue are based on the ideal deterministic network model, we study the data collection problem for practical probabilistic WSNs. We design a cell-based path scheduling algorithm and …

Maintaining Integrity Constraints In Semantic Web, Ming Fang

Computer Science Dissertations

As an expressive knowledge representation language for Semantic Web, Web Ontology Language (OWL) plays an important role in areas like science and commerce. The problem of maintaining integrity constraints arises because OWL employs the Open World Assumption (OWA) as well as the Non-Unique Name Assumption (NUNA). These assumptions are typically suitable for representing knowledge distributed across the Web, where the complete knowledge about a domain cannot be assumed, but make it challenging to use OWL itself for closed world integrity constraint validation. Integrity constraints (ICs) on ontologies have to be enforced; otherwise conflicting results would be derivable from the same …

Scientific High Performance Computing (Hpc) Applications On The Azure Cloud Platform, Dinesh Agarwal

Computer Science Dissertations

Cloud computing is emerging as a promising platform for compute and data intensive scientific applications. Thanks to the on-demand elastic provisioning capabilities, cloud computing has instigated curiosity among researchers from a wide range of disciplines. However, even though many vendors have rolled out their commercial cloud infrastructures, the service offerings are usually only best-effort based without any performance guarantees. Utilization of these resources will be questionable if it can not meet the performance expectations of deployed applications. Additionally, the lack of the familiar development tools hamper the productivity of eScience developers to write robust scientific high performance computing (HPC) applications. …

Collaborative Communication And Storage In Energy-Synchronized Sensor Networks, Mingsen Xu

Computer Science Dissertations

In a battery-less sensor network, all the operation of sensor nodes are strictly constrained by and synchronized with the fluctuations of harvested energy, causing nodes to be disruptive from network and hence unstable network connectivity. Such wireless sensor network is named as energy-synchronized sensor networks. The unpredictable network disruptions and challenging communication environments make the traditional communication protocols inefficient and require a new paradigm-shift in design. In this thesis, I propose a set of algorithms on collaborative data communication and storage for energy-synchronized sensor networks. The solutions are based on erasure codes and probabilistic network codings. The proposed set of …