Daily fantasy sports (DFS) has grown in popularity with millions of participants throughout the world. However, studies have shown that most profits from DFS contests are won by only a small percentage of players. This thesis addresses the challenges faced by DFS participants by evaluating sources that provide player projections for NBA DFS contests and by developing machine learning models that produce competitive player projections. External sources are evaluated by constructing daily lineups based on the projections offered and evaluating those lineups in the context of all potential lineups, as well as those submitted by participants in competitive FanDuel DFS tournaments. Lineups produced by the machine learning models are also evaluated in the same manner. This work experiments with several machine learning techniques including automated machine learning and notes the top model developed was successful in 48% of all FanDuel NBA DFS tournaments and 51% of single-entry tournaments over a two-month period, surpassing the top external source evaluated by 9 percentage points and 10 percentage points, respectively.

Today’s online content increases at an alarmingly rate which exceeds users’ ability to consume such content. Modern search techniques allow users to enter keyword queries to find content they wish to see. However, such techniques break down when users freely browse the internet without knowing exactly what they want. Users may have to invest an unnecessarily long time reading content to see if they are interested in it. Automatic text summarization helps relieve this problem by creating synopses that significantly reduce the text while preserving the key points. Steffen Lyngbaek created the SPORK summarization pipeline to solve the content overload in Reddit comment threads. Lyngbaek adapted the Opinosis graph model for extractive summarization and combined it with agglomerative hierarchical clustering and the Smith-Waterman algorithm to perform multi-document summarization on Reddit comments.This thesis presents WHISK as a pipeline for general multi-document text summarization based on SPORK. A generic data model in WHISK allows creating new drivers for different platforms to work with the pipeline. In addition to the existing Opinosis graph model adapted in SPORK, WHISK introduces two simplified graph models for the pipeline. The simplified models removes unnecessary restrictions inherited from Opinosis graph’s abstractive summarization origins. Performance measurements and a study with Digital Democracy compare the two new graph models against the Opinosis graph model. Additionally, the study evaluates WHISK’s ability to generate pull quotes from political discussions as summaries.

Denial-of-Service (DoS) attacks are aimed at shutting a machine or network down to block users from accessing it. These attacks can be difficult to detect and can cost millions in damages or lost earnings. Since the first DoS attack occurred in 1999, the way DoS attacks have been launched has become more complicated, making them more elusive and harder to detect. The first step to detect and mitigate a DoS attack is for a system to identify the malicious traffic. In this experiment, we aim to identify the malicious traffic within ten seconds. To do this the project was divided into 3 phases: data collection, feature extraction and construction of classification. The first phase was to collect malicious and legitimate data using Wireshark. The second phase of the project was to convert the PCAP files into features that are meaningful and easy to read. The third phase of the project is the construction of classification models. We used the Naïve Bayes and decision tree classification models to identify malicious traffic data and differentiate it from legitimate traffic data. This approach yielded an 𝐹1 score average of 92% in detecting DoS attacks and an 𝐹1 𝑠𝑐𝑜𝑟𝑒 accuracy range of 37% to 71% to accurately determine the intensity of the DoS attack, a reasonable accuracy for this problem. These results show that it is possible to not only detect DoS attacks, but also, to determine the intensity of such attacks with a reasonable accuracy.

<p>Localization of speech onsets to determine onset latencies is a complicated problem with as many different methods for finding them as there are different areas which use such measurements. A majority of research performed in cognition uses a standard amplitude threshold voice key for estimating the speech onset latencies but a number of studies have shown that this method is incredibly inaccurate and can bias data or produce contradictory results. A number of alternative methods based on modifications to traditional voice-keys have been proposed to deal with the inconsistency although still show a number of deficiencies. Previous work has suggested that switching from the amplitude domain of a signal to the frequency domain a number of the issues present with voice keys can be overcome and when used in conjunction with a number of highly sensitive heuristics highly accurate onset latencies can be produced reliably under ideal conditions. This research is refined and paired with a new user interface to improve the ease of use and increase the adoption rate of this type of analysis. Recent work in the telecommunications industry also suggests that wavelet-based algorithms in conjunction with the Teager Energy Operator (TEO) can accurately detect speech even in the presence of noise. Four wavelet-based methods are investigated and tested; a simple wavelet transform test, and three methods using wavelet-packet transforms in conjunction with the TEO. Although these methods do not perform very well compared to traditional methods a number of potential issues with the implementation are discussed.</p> / Master of Science (MSc)

The dissertation proposes, presents and analyzes a new design pattern, the Mutable Class pattern, to support the processing of large-scale heterogeneous data models with multiple families of algorithms. Handling data-algorithm associations represents an important topic across a variety of application domains. As a result, it has been addressed by multiple approaches, including the Visitor pattern and the aspect-oriented programming (AOP) paradigm. Existing solutions, however, bring additional constraints and issues. For example, the Visitor pattern freezes the class hierarchies of application models and the AOP-based projects, such as Spring AOP, introduce significant overhead for processing large-scale models with fine-grain objects. The Mutable Class pattern addresses the limitations of these solutions by providing an alternative approach designed after the Class model of the UML specification. Technically, it extends a data model class with a class mutator supporting the interchangeability of operations. Design patterns represent reusable solutions to recurring problems. According to the design pattern methodology, the definition of these solutions encompasses multiple topics, such as the problem and applicability, structure, collaborations among participants, consequences, implementation aspects, and relation with other patterns. The dissertation provides a formal description of the Mutable Class pattern for processing heterogeneous tree-based models and elaborates on it with a comprehensive analysis in the context of several applications and alternative solutions. Particularly, the commonality of the problem and reusability of this approach is demonstrated and evaluated within two application domains: computational accelerator physics and compiler construction. Furthermore, as a core part of the Unified Accelerator Library (UAL) framework, the scalability boundary of the pattern has been challenged and explored with different categories of application architectures and computational infrastructures including distributed three-tier systems. The Mutable Class pattern targets a common problem arising from software engineering: the evolution of type systems and associated algorithms. Future research includes applying this design pattern in other contexts, such as heterogeneous information networks and large-scale processing platforms, and examining variations and alternative design patterns for solving related classes of problems.

Public transportation plays an integral part in a city's development, but transportation professionals disagree about whether it is feasible to increase the capacity of public transportation systems at a reasonable cost; and if it is, how. This study develops an agent based model that aims to answer this question and provide a framework to compare the effects of improvements in different aspects of the public transportation service. The results of this study show that it is possible to increase ridership enough to compensate for the increased operational cost, but only in certain circumstances. Interesting phenomenon that might have showed up in the real world arose in this model and are worth further investigation.

The field of biometrics has grown significantly in the past decade due to an increase in interest from law enforcement. Law enforcement officials are interested in adding tattoos alongside irises and fingerprints to their toolbox of biometrics. They often use these biometrics to aid in the identification of victims and suspects. Like facial recognition, tattoos have seen a spike in attention over the past few years. Tattoos, however, have not received as much attention by researchers. This lack of attention towards tattoos stems from the difficulty inherent in matching these tattoos. Such difficulties include image quality, affine transformation, warping of tattoos around the body, and in some cases, excessive body hair covering the tattoo. We will utilize context-based image retrieval to find a tattoo in a database which means using one image to query against a database in order to find similar tattoos. We will focus specifically on the keypoint detection process in computer vision. In addition, we are interested in finding not just exact matches but also similar tattoos. We will conclude that the ORB detector pulls the most relevant features and thus is the best chance for yielding an accurate result from content-based image retrieval for tattoos. However, we will also show that even ORB will not work on its own in a content-based image retrieval system. Other processes will have to be involved in order to return accurate matches. We will give recommendations on next-steps to create a better tattoo retrieval system.

Digital forensics is the practice of trained investigators gathering and analyzing evidence from digital devices such as computers and smart phones. On these digital devices, it is possible to change the time on the device for a purpose other than what is intended. Currently there are no documented techniques to determine when this occurs. This research seeks to prove out a technique for determining when the time has been changed on forensic disk image by analyzing the log files found on the image. Out of this research a tool is created to perform this analysis in automated fashion. This tool is TADpole, a command line program that analyzes the log files on a disk image and determines if a timeline anomaly has occurred.

This study explores the use of neural networks as universal models for classifying file fragments. This approach differs from previous work in its lossless feature representation, with fragments’ bits as direct input, and its use of feedforward, recurrent, and convolutional networks as classifiers, whereas previous work has only tested feedforward networks. Due to the study’s exploratory nature, the models were not directly evaluated in a practical setting; rather, easily reproducible experiments were performed to attempt to answer the initial question of whether this approach is worthwhile to pursue further, especially due to its high computational cost. The experiments tested classification of fragments of homogeneous file types as an idealized case, rather than using a realistic set of types, because the types of interest are highly application-dependent. The recurrent networks achieved 98 percent accuracy in distinguishing 4 file types, suggesting that this approach may be capable of yielding models with sufficient performance for practical applications. The potential applications depend mainly on the model performance gains achievable by future work but include binary mapping, deep packet inspection, and file carving.

Videoconferencing is increasingly becoming a trend worldwide in applications where clients need to access lectures, meeting proceedings, communicating with family and friends etc. It provides a platform enabling the visual, audio and video communication between clients. The aim of this project is to utilize the open source Java software to build a desktop application enabling communication between clients. When a user needs to transfer a secured file, it’s unsafe to send it using social networking sites because of lack of security. So, with the “Multi-Way Communication System” (MWCS) we resolve some security issues. The MWCS is a highly secure way for file transfer, text and video conferencing.