-
Prospective Students
-
Faculty and Staff
-
Current Students
-
Alumni and Friends
To address the challenges of granular emotion detection in social media text (EMDISM), I have investigated ensemble approaches that combine a variety of individual classifiers to address tradeoffs in performance. This involved first investigating EMDISM performance for individual traditional machine learning (ML), deep learning (DL), and transformer learning (TL) classifiers. Based on this analysis, the second stage investigated the creation of ensembles of the most accurate classifiers across these general classes which offer comparatively improved performance. I provide results and analysis for each classifier I considered as well as the most accurate ensembles I created from the most accurate singleton classifiers. Results show that the proposed ensemble approaches improve upon the state of the art for average accuracy, weighted precision, weighted recall, and weighted f-measure as compared to the most accurate single classifier for EMDISM.
In recent years there has been a significant increase in the prevalence of mental illness among millennials (White-Cummings, 2017). However, there is still a significantly lower rate of Black millennials, specifically Black men, utilizing mental health services compared to other marginalized groups (Cadaret & Speight, 2018). Black men have reportedly have a higher prevalence of mental illness with little to no treatment engagement, which has been linked to the increasingly high rates of suicide. Black men and their lack of mental health treatment seeking has become an increasingly popular topic in scholarly literature, yet the research is still scarce thus far. The purpose of this study was to explore the influence of social constructs on millennial Black men’s decisions about seeking mental health treatment through the lens of Critical Race Theory (CRT), Black Critical Theory (BlackCrit) and Black Masculinity. Based on past reported themes, Black Masculinity, CRT, and BlackCrit were utilized as a multidimensional framework for this critical phenomenology qualitative study. The researcher used semi-structured interviews to investigate the experiences of 16 participants who identified as millennial Black men that had considered seeking mental health treatment regardless of their decision to seek help or not. Following a modified version of Moustakas (1994) phenomenological analysis, results indicated three themes Racialized Gendered Socialization, Cultural Distrust, and Invisibility. All themes were related to racial and masculine factors. Implications and recommendations are provided for future research and improving advocacy efforts to engage more Black men in mental health treatment.
Using employee referral programs is generally considered a best practice for organizations seeking top quality talent. However, research on whether or not these programs result in positive outcomes equally for all applicants is mixed. To data, most research examining employee referral programs focuses on how status characteristics (such as race and gender) of applicants can result in unequal outcomes (such as being hired or promoted) for applicants with different identities. Little is known, however, about the influence of referring employee’s status characteristics during hiring processes and whether or not decision makers’ biases toward certain referring employees may lead to different hiring process outcomes for the applicants they refer. Using Status Characteristics Theory and the theory of Status Beliefs Transfer, hypotheses were tested regarding how status characteristics of referring employees, namely race and gender, might lead to a transfer of evaluators’ status beliefs from the referring employee to the applicant and affect subsequent applicant evaluations. Four hundred and thirty-seven U.S. individuals with hiring experience served as participants for an online resume evaluation experiment where the only difference between resumes was the name of the referring employee noted at the top of the document. Referring employee names were selected via pre-test to signal the referrer was either a white man, black man, white woman, or black women. Results of quantitative analyses revealed a positive statistically significant difference in average ratings of competence, recommendations for interviews, and starting salary between referred and non-referred applicants, with participants rating referred applicants more favorably. In addition, a positive statistically significant effect of race, but not gender, was found in average ratings of competence, commitment, interview recommendations, and salary recommendations for black referring compared to white referring employees. Additional qualitative thematic analysis of open response data describing rationale for participant ratings revealed additional intersectional evaluative differences among applicants referred by employees with different race/gender statuses. Taken together, and viewed through the lens of intersectional theories, findings suggest evaluations of applicants may have been influenced by a status beliefs transfer process whereby the intersectional status characteristics of referring employees were transferred onto and used to evaluate the applicants they referred. Implications for theory, practice, and future research are discussed.
Connected and autonomous vehicle (CAV) technologies could significantly change the car-following behaviors and affect the performance of the intersection systems. As it is expected to have a long transition time during which human driven vehicles (HDVs) and CAVs will coexist, it is important to investigate the impacts of CAVs on the intersection systems under different market penetration rates (MPRs). Also, the currently used Highway Capacity Manual does not consider the impacts of CAVs when calculating the intersection capacity. Though highly needed, a new guideline for estimating the intersection capacity under different MPRs of CAVs is becoming a critical issue for transportation planners and engineers. Furthermore, combining the intersection traffic signal control (TSC) systems with deep reinforcement learning (DRL) provides a new potential solution to improve the efficiency, safety, and sustainability of the intersection system. However, the training procedure of the DRL TSC system requires large samples and takes a long time to converge. Furthermore, it is common to have several intersections along corridors or in networks. A single DRL agent is unable to control several intersections as this may result in exponential explosion in the action space. Hence, a modification of the DRL TSC framework to improve the training efficiency and a multi-agent control framework to control several intersections are needed.
To better prepare and guide both intersection planning and operations under different MPRs of CAVs and traffic demands, this dissertation provides an intensive evaluation of the impacts of CAVs in several signal intersection systems, as well as an in-depth analysis on intersection capacity adjustments that consider varying MPRs of CAVs. Also, a transfer-based DRL TSC framework is proposed and tested at different MPRs of CAVs and traffic demand levels. A multi-agent DRL TSC with shared traffic states between downstream and upstream intersections is investigated in a corridor. It is concluded that 100% MPR of CAVs can increase the saturation flow rate of the through-only lane by 126.8%. Meanwhile, transfer-based models could significantly improve training efficiency and model performance. The multi-agent DRL TSC also enables coordination between intersections. The insights of this research should be helpful and valuable to transportation researchers and traffic engineers in calculating intersection capacity, designing intelligent intersections, improving intersection efficiency, and implementing DRL-controlled traffic signals under the mixed flow with CAVs.
Connected and autonomous vehicles (CAVs) are a type of emerging technology that has promising potentials in improving many aspects of the existing transportation infrastructure, including operations, safety, and the environment. With the capability of traveling on the roads with shorter headways and more stable speeds, CAVs can yield a larger road capacity compared to human-driven vehicles (HDVs). Additionally, since the CAVs run on the roads with the guidance of computers or algorithms, accidents caused by errors from human drivers may be prevented, which can greatly reduce significant economic and societal losses. Less speed fluctuations are also beneficial to decrease emissions and contribute to the environment.
Thanks to the rapid development of computer science and communication technology, CAVs have evolved from theoretical experiments in academic labs to reliable products by commercial companies. Since both academic and industrial professionals have high expectations for CAVs, many studies have been conducted to explore and identify the impacts of CAV technologies on the transportation performances in many scenarios. These scenarios included conventional intersections, highway segments, on/off ramps, and roundabouts. Through extensive investigations on CAVs in different scenarios, it can be concluded that CAVs can perform better overall than HDVs. Nevertheless, it has also been found that the performances of CAVs are affected by many factors such as communication range, acceleration capabilities, and market penetration rates. Improvement in operational performance has been confirmed by existing studies when the market penetration of CAVs reaches a certain rate.
Superstreet is one of the innovative intersection designs and was proposed to alleviate the road congestion especially where unbalanced traffic volumes from main street and minor street exist. Superstreets have been successfully implemented in numerous states. Nevertheless, how CAVs would affect the performances of superstreets has not been explored, even to a minimum extent. This research is designed to investigate how CAVs with different technologies perform in the environment of superstreets. To be specific, the following questions will be answered: (1) at what market penetration rate CAVs would bring benefits towards operational performances; (2) at what extent CAVs would bring benefits towards operational performances of superstreets; (3) how the impact of CAVs on the operational performance would vary across different traffic scales and market penetration rates.
To achieve the research goals, models for CAV platooning, trajectory planning, and signal optimization have been developed, respectively. The effects of these models are tested respectively in a simulation environment where relevant traffic measures are extracted to evaluate the performances. The finding of this research may also be applied to other innovative intersection designs which have similar geometric characteristics and traffic patterns.
ABSTRACT
SHOHREH SHADALOU. Dynamic Illumination Systems using Freeform Optics. (Under the direction of DR. THOMAS J. SULESKI)
Illumination systems that can create light patterns of varying sizes or shapes with high efficiency and uniformity are advantageous for a range of applications, including lighting, augmented/virtual reality, laser-based manufacturing, medicine/dermatology, and lithography. Previous approaches for continuous variable illumination have utilized longitudinal movement of the source or other optical components along the optical axis, which increases both system size and light pattern non uniformity. Liquid lenses with adjustable membranes have also been used for tunable illumination, but leakage and manufacturing complexity can be significant issues. Thus, new approaches that enable dynamically tunable illumination patterns in compact, robust packages are of interest.
Recent advances in design, production and metrology have enabled the use of freeform surfaces in a wide range of optical imaging applications. As one example, the Alvarez lens consists of a pair of cubic freeform surfaces that enable variable focal length with small lateral displacements between the two elements. Complex freeform surfaces are also regularly used in static illumination systems such as automotive headlights and luminaires.
The primary objectives of this dissertation are to explore and characterize dynamic freeform optical systems enabling continuously variable illumination. Results are addressed through three articles. The first article introduces the use of arrays of freeform Alvarez lenses with LED sources to enable tunable illumination. The second article builds from this work to present the design, manufacturing, and characterization of a compact tunable illumination system. The third article introduces a general design method using freeform optics to enable variable optical illumination between two arbitrary boundary conditions. These three articles demonstrate the methods and utility of freeform optics for dynamic illumination systems.
Employee overqualification is becoming increasingly relevant in a post-pandemic world. While there have been theoretical advancements in the overqualification literature, several methodological issues remain unresolved. Specifically, the conceptualization and operationalization of perceived overqualification (POQ) are often not aligned. To date, the perception of overqualification is not yet fully understood. Thus, the main goal of this dissertation is to address these methodological limitations. In Study 1, I refined the scope of POQ by offering an explicit construct conceptualization grounded in person-job fit theory and developed a new scale to measure the multidimensional construct. In Study 2, I validated the psychometric properties of the Perceived Overqualification at Work Scale (POQWS) and explored the relationship of POQ with various work-related outcomes. Taking a person-centric approach, I used latent profile analyses (LPA) to identify different profiles of overqualified employees in Study 3 based on the POQWS dimensions. This study is the first to examine the process by which patterns of variables are identified in POQ profiles and how these combinations differentially relate to outcomes. Results from a series of exploratory and confirmatory factor analyses clearly supported a four-factor model. In the subsequent study, four distinct profiles emerged from the latent profile analyses. One-way analyses of variance (ANOVA) provided further criterion-related validity evidence for these four profiles. Taken together, the findings from this dissertation lay the grounds for future person-centered research.
The construct of “customer satisfaction” has been used for several decades in marketing to achieve outcomes such as customer loyalty, word-of-mouth communication, resistance to competition, and customer equity. Recent research, however, has indicated little to no correlation between customer satisfaction and many of these outcomes. A more recent marketing construct is “customer delight,” where affective bonds and positive associations are the foundations for customer relationships. While customer delight has numerous advantages, an important limitation is that it can only be used with certain types of products and consumption situations.
This study introduces the academic construct of “customer success,” an objective tool that could redefine customer relationships, and define it as an objective and mathematically based strategic process to maximize customer-desired outcomes. A long-term customer success strategy is customer-driven and designed to be mutually beneficial to both an organization and its customers. While the construct of customer success has been sporadically used by practitioners in the past, the use of the term has often been arbitrary, and the construct has never been precisely defined.
First, drawing on the reverse logic framework (RLF) of relationship marketing, the customer valuation model, and return on relationships (ROR), this study will use Hunt’s indigenous theory, inductive realist approach to help build the initial theoretical framework for the construct of customer success. Then, this study uses this construct in a government-to-customer (G2C) market scenario to test a series of hypotheses to evaluate government-achieved customer success for COVID-19 pandemic response outcomes. This study will conclude with theoretical and managerial research contributions and provide directions for future research.
Resource-hungry applications play a very important role in people's daily lives, such as real-time video streaming applications and mobile augmented reality applications.
However, there are several challenges to satisfy the user Quality-of-Experience (QoE) requirements of resource-hungry applications. First, these applications usually require a vast amount of network bandwidth resources to support the data communication of different functionalities. However, only limited network bandwidth resources can be assigned to these applications which leads to long network latency and poor user QoE. In addition, artificial intelligent (AI) and machine learning (ML) models are widely adopted in these applications which significantly increases the computation complexity of these applications. Because of the limited computing resource on mobile devices, computation-intensive tasks are offloaded to edge servers located at the edge of the core network. However, additional network latency and bandwidth usage are introduced which may degrade user QoE. In this dissertation, the characteristics of popular resource-hungry applications are first analyzed. Then, based on the analyzed characteristics, we propose several specific ally designed algorithms to enhance the performance of several popular resource-hungry applications.
