Cycling is an underutilized mode of transportation in cities across North America. Numerous factors contribute to low ridership levels, but a key deterrent to cycling is concern for personal safety. In an effort to increase cycling mode share, many cities are investing in cycling infrastructure, with several cities constructing connected bicycle networks. Monitoring the impact of new infrastructure is important for accountability to citizens and to encourage political will for future investments in cycling facilities. A lack of spatially continuous ridership data and methodological challenges have limited monitoring and evaluation of the impacts of infrastructure changes. The goal of our research was to demonstrate spatially explicit approaches for monitoring city-wide changes in patterns of safety and ridership following improvements to cycling infrastructure. To meet our goal, our first analysis demonstrated a method for monitoring changes in the spatial-temporal distribution of cycling incidents across a city. We compared planar versus network constrained kernel density estimation for visualizing cycling incident intensity across the street network of Vancouver, Canada using cycling incidents reported to the Insurance Corporation of British Columbia. Next, we applied a change detection algorithm to detect statistically significant change between maps of kernel density estimates. The utility of the network kernel density change detection method is demonstrated through a case study in the city of Vancouver, Canada where we compare cycling incident densities following construction of two cycle tracks in the downtown core. The methods developed and demonstrated for this study provide city planners, transportation engineers and researchers a means of monitoring city-wide changes in the patterns of cycling incidents following enhancements to cycling infrastructure. Our second analysis demonstrated how network constrained spatial analysis methods can be applied to emerging sources of crowdsourced cycling data to monitor city-wide changes in patterns of ridership. We used network constrained global and local measures of spatial autocorrelation, applied to crowdsourced ridership data from Strava, to examine changes in ridership patterns across Ottawa-Gatineau, Canada, following installation and closures of cycling infrastructure. City planners, transportation engineers and researchers can use the methods outlined here to monitor city-wide changes in ridership patterns following investment in cycling infrastructure or other changes to the transportation network. Through this thesis we help overcome the challenges associated with monitoring the impact of infrastructure changes on ridership and cycling safety. We demonstrated how network constrained spatial analysis methods can be applied to officially reported cycling incident data to identify changes in the spatial-temporal distribution of cycling safety across a transportation network. We also demonstrated how network appropriate spatial analysis techniques can be applied to large, emerging crowdsourced cycling datasets to monitor changes in patterns of ridership. These methods enhance our understanding of the city-wide impact of infrastructure changes on cycling safety and ridership patterns. / Graduate

The first chapter studies mutual funds. I model intraquarter trading and use a genetic algorithm to estimate the trade pattern that is most consistent with the fund's daily reported returns. I validate the model empirically on a sample of institutional trades from Ancerno and I confirm that the method more accurately predicts daily holdings when compared to existing naive assumptions. Further, my method is substantially more accurate in classifying a fund's tendency to supply liquidity, and this increased precision has important implications for identifying superior performing funds. Specifically, a long-short strategy based on the model's liquidity provision measures earns significant abnormal returns, while a similar strategy that relies on quarterly holdings does not exhibit any outperformance. The second chapter studies investment research. We find evidence that crowdsourced investment research facilitates informed trading by retail investors and improves firm liquidity. Specifically, retail order imbalances are strongly correlated with the sentiment of Seeking Alpha articles, and the ability of retail order imbalances to predict returns is roughly twice as large on research article days. In addition, firms with exogenous reductions in Seeking Alpha coverage experience increases in bid-ask spreads and price impact, with the effect being stronger for firms with high retail ownership. Our findings suggest that technological innovations have helped democratize access to investment research with important implications for firm liquidity.

abstract: Factors that explain human mobility and active transportation include built environment and infrastructure features, though few studies incorporate specific geographic detail into examinations of mobility. Little is understood, for example, about the specific paths people take in urban areas or the influence of neighborhoods on their activity. Detailed analysis of human activity has been limited by the sampling strategies employed by conventional data sources. New crowdsourced datasets, or data gathered from smartphone applications, present an opportunity to examine factors that influence human activity in ways that have not been possible before; they typically contain more detail and are gathered more frequently than conventional sources. Questions remain, however, about the utility and representativeness of crowdsourced data. The overarching aim of this dissertation research is to identify how crowdsourced data can be used to better understand human mobility. Bicycling activity is used as a case study to examine human mobility because smartphone apps aimed at collecting bicycle routes are readily available and bicycling is under studied in comparison to other modes. The research herein aimed to contribute to the knowledge base on crowdsourced data and human mobility in three ways. First, the research examines how conventional (e.g., counts, travel surveys) and crowdsourced data correspond in representing bicycling activity. Results identified where the data correspond and differ significantly, which has implications for using crowdsourced data for planning and policy decisions. Second, the research examined the factors that influence cycling activity generated by smartphone cycling apps. The best predictors of activity were median weekly rent, percentage of residential land, and the number of people using two or more modes to commute in an area. Finally, the third part of the dissertation seeks to understand the impact of bicycle lanes and bicycle ridership on residential housing prices. Results confirmed that bicycle lanes in the neighborhood of a home positively influence sale prices, though ridership was marginally related to house price. This research demonstrates that knowledge obtained through crowdsourced data informs us about smaller geographic areas and details on where people bicycle, who uses bicycles, and the impact of the built environment on bicycling activity. / Dissertation/Thesis / Doctoral Dissertation Geography 2018

Intelligent transportation systems (ITS) are essential tools for traffic planning, analysis, and forecasting that can utilize the huge amount of traffic data available nowadays. In this work, we aggregated detailed traffic flow sensor data, Waze reports, OpenStreetMap (OSM) features, and weather data, from California Bay Area for 6 months. Using that data, we studied three novel ITS applications using convolutional neural networks (CNNs) and recurrent neural networks (RNNs). The first experiment is an analysis of the relation between roadway shapes and accident occurrence, where results show that the speed limit and number of lanes are significant predictors for major accidents on highways. The second experiment presents a novel method for forecasting congestion severity using crowdsourced data only (Waze, OSM, and weather), without the need for traffic sensor data. The third experiment studies the improvement of traffic flow forecasting using accidents, number of lanes, weather, and time-related features, where results show significant performance improvements when the additional features where used.

Yes / The domain of cultural variations in interpersonal communication is becoming increasingly important in various areas, including human-human interaction (e.g. business settings) and humancomputer interaction (e.g. during simulations, or with social robots). User generated content (UGC) in social media can provide an invaluable source of culturally diverse viewpoints for supporting the understanding of cultural variations. However, discovering and organizing UGC is notoriously challenging and laborious for humans, especially in ill-defined domains such as culture. This calls for computational approaches to automate the UGC sensemaking process by using tagging, linking and exploring. Semantic technologies allow automated structuring and qualitative analysis of UGC, but are dependent on the availability of an ontology representing the main concepts in a specific domain. For the domain of cultural variations in interpersonal communication, no ontological model exists. This paper presents the first such ontological model, called AMOn+, which defines cultural variations and enables tagging culture-related mentions in textual content. AMOn+ is designed based on a novel interdisciplinary approach that combines theoretical models of culture with crowdsourced knowledge (DBpedia). An evaluation of AMOn+ demonstrated its fitness-for-purpose regarding domain coverage for annotating culture-related concepts mentioned in text corpora. This ontology can underpin computational models for making sense of UGC.

The successful crowdsourcing model and gamification design of Stack Overflow (SO) Q&A platform have attracted many programmers to ask and answer technical questions, regardless of their level of expertise. Researchers have recently found evidence of security vulnerable code snippets being possibly copied from SO to production software. This inspired us to study how reliable is SO in providing secure coding suggestions. In this project, we automatically extracted answer posts related to Java security APIs from the entire SO site. Then based on the known misuses of these APIs, we manually labeled each extracted code snippets as secure or insecure. In total, we extracted 953 groups of code snippets in terms of their similarity detected by clone detection tools, which corresponds to 785 secure answer posts and 644 insecure answer posts. Compared with secure answers, counter-intuitively, insecure answers has higher view counts (36,508 vs. 18,713), higher score (14 vs. 5), more duplicates (3.8 vs. 3.0) on average. We also found that 34% of answers provided by the so-called trusted users who have administrative privileges are insecure. Our finding reveals that there are comparable numbers of secure and insecure answers. Users cannot rely on community feedback to differentiate secure answers from insecure answers either. Therefore, solutions need to be developed beyond the current mechanism of SO or on the utilization of SO in security-sensitive software development. / Master of Science / Stack Overflow (SO), the most popular question and answer platform for programmers today, has accumulated and continues accumulating tremendous question and answer posts since its launch a decade ago. Contributed by numerous users all over the world, these posts are a type of crowdsourced knowledge. In the past few years, they have been the main reference source for software developers. Studies have shown that code snippets in answer posts are copied into production software. This is a dangerous sign because the code snippets contributed by SO users are not guaranteed to be secure implementations of critical functions, such as transferring sensitive information on the internet. In this project, we conducted a comprehensive study on answer posts related to Java security APIs. By labeling code snippets as secure or insecure, contrasting their distributions over associated attributes such as post score and user reputation, we found that there are a significant number of insecure answers (644 insecure vs 785 secure in our study) on Stack Overflow. Our statistical analysis also revealed the infeasibility of differentiating between secure and insecure posts leveraging the current community feedback system (eg. voting) of Stack Overflow.

abstract: Dynamic spectrum access (DSA) has great potential to address worldwide spectrum shortage by enhancing spectrum efficiency. It allows unlicensed secondary users to access the under-utilized spectrum when the primary users are not transmitting. On the other hand, the open wireless medium subjects DSA systems to various security and privacy issues, which might hinder the practical deployment. This dissertation consists of two parts to discuss the potential challenges and solutions. The first part consists of three chapters, with a focus on secondary-user authentication. Chapter One gives an overview of the challenges and existing solutions in spectrum-misuse detection. Chapter Two presents SpecGuard, the first crowdsourced spectrum-misuse detection framework for DSA systems. In SpecGuard, three novel schemes are proposed for embedding and detecting a spectrum permit at the physical layer. Chapter Three proposes SafeDSA, a novel PHY-based scheme utilizing temporal features for authenticating secondary users. In SafeDSA, the secondary user embeds his spectrum authorization into the cyclic prefix of each physical-layer symbol, which can be detected and authenticated by a verifier. The second part also consists of three chapters, with a focus on crowdsourced spectrum sensing (CSS) with privacy consideration. CSS allows a spectrum sensing provider (SSP) to outsource the spectrum sensing to distributed mobile users. Without strong incentives and location-privacy protection in place, however, mobile users are reluctant to act as crowdsourcing workers for spectrum-sensing tasks. Chapter Four gives an overview of the challenges and existing solutions. Chapter Five presents PriCSS, where the SSP selects participants based on the exponential mechanism such that the participants' sensing cost, associated with their locations, are privacy-preserved. Chapter Six further proposes DPSense, a framework that allows the honest-but-curious SSP to select mobile users for executing spatiotemporal spectrum-sensing tasks without violating the location privacy of mobile users. By collecting perturbed location traces with differential privacy guarantee from participants, the SSP assigns spectrum-sensing tasks to participants with the consideration of both spatial and temporal factors. Through theoretical analysis and simulations, the efficacy and effectiveness of the proposed schemes are validated. / Dissertation/Thesis / Doctoral Dissertation Electrical Engineering 2017

This thesis analyzes the end-user quality of service (QoS) in cellular mobile networks using device-side measurements. Quality of service in a wireless network is a significant factor in determining a user's satisfaction. Customers' perception of poor QoS is one of the core sources of customer churn for telecommunications companies. A core focus of this work is on assessing how user density impacts QoS within cellular networks. Kernel density estimation is used to produce user density estimates for high, medium, and low density areas. The QoS distributions are then compared across these areas. The k-sample Anderson-Darling test is used to determine the degree to which user densities vary over time. In general, it is shown that users in higher density areas tend to experience overall lower QoS levels than those in lower density areas, even though these higher density areas service more subscribers. The conducted analyses highlight the value of mobile device-side QoS measurements in augmenting traditional network-side QoS measurements. / Graduate

This thesis aims to explore and describe the content diversity of Siri’s search results in the polarized context of US politics. To do so, a crowdsourced audit was conducted. A diverse sample of 134 US-based Siri users between the ages of 18-64 performed five identical queries about the politically controversial issues of gun laws, immigration, the death penalty, taxes and abortion. The data were viewed through a theoretical framework using the concepts of algorithmic bias and media-centric fragmentation. The results suggest that Siri’s search algorithm produces a long tail distribution of search results: Forty-two percent of the participants received the six most frequent answers, while 22% of the users received unique answers. These statistics indicate that Siri’s search algorithm causes moderate concentration and low fragmentation. The age and, surprisingly, the political orientation of users, do not seem to be driving either concentration or fragmentation. However, the users' gender and location appears to cause low concentration. The finding that Siri’s search algorithm produces a long tail of replies challenges previous research on the content diversity of search results, which found no evidence of fragmentation. However, due to the limited scope of this study, these findings cannot be generalized to a larger population. Further research is needed to support or refute them.

In online social networking sites, gaining popularity has become important. The more popular a company is, the more profits it can make. A way to measure a company's popularity is to check how many likes it has (e.g., the company's number of likes in Facebook). To instantly and artificially increase the number of likes, some companies and business people began hiring crowd workers (aka fake likers) who send likes to a targeted page and earn money. Unfortunately, little is known about characteristics of the fake likers and how to identify them. To uncover fake likers in online social networks, in this work we (i) collect profiles of fake likers and legitimate likers by using linkage and honeypot approaches, (ii) analyze characteristics of fake likers and legitimate likers, (iii) propose and develop a fake liker detection approach, and (iv) thoroughly evaluate its performance against three baseline methods and under two attack models. Our experimental results show that our cassification model significantly outperformed the baseline methods, achieving 87.1% accuracy and 0.1 false positive rate and 0.14 false negative rate.