22505 research outputs found
Sort by
Predicting Student Dropout Risk using Machine Learning
Get PDFStudent dropout remains a persistent challenge in higher education, undermining institutional performance, reducing workforce preparedness, and limiting students’ academic and economic opportunities. Accurately identifying students at risk of attrition is complex, due to the interplay of academic, financial, and behavioral factors. This thesis addresses this challenge by applying a combined machine learning framework—integrating both unsupervised and supervised techniques—to predict student dropout using structured, first-year academic and financial data. The study utilizes a comprehensive dataset of 4,424 undergraduate student records from a European higher education institution, covering ten academic years and comprising 35 variables related to academic performance, enrollment behavior, and financial engagement. Clustering techniques were employed to group students by engagement profiles, while classification models—including Random Forest, XGBoost, and a soft voting ensemble—were trained to predict final academic outcomes: Dropout, Enrolled, or Graduate. Feature engineering was conducted in two phases, with both semester-averaged metrics and advanced derived indicators used to enhance model performance. Findings show that academic approvals, grades, and tuition fee status are the most influential predictors of student outcomes. Unsupervised clustering revealed behaviorally distinct groups with statistically significant dropout risks, though these clusters did not translate effectively into predictive labels. Supervised models, particularly tuned XGBoost and ensemble classifiers, achieved high performance in binary classification tasks (balanced accuracy ¿ 0.91, AUC ¿ 0.95), confirming that dropout risk can be reliably predicted from early academic records. However, multiclass classification performance declined, especially for the transitional “Enrolled” category, highlighting the limitations of static early-year data in capturing more ambiguous student states. This research contributes to the literature by demonstrating the strengths and constraints of interpretable machine learning in modeling student success. It also offers actionable insights for academic institutions, such as prioritizing interventions for students with early signs of disengagement and financial instability. Methodologically, the study highlights opportunities for future work to explore hybrid clustering-classification models, apply soft clustering techniques, and evaluate deep learning models for benchmarking purposes. While complex models may lack interpretability, they can serve as useful baselines to understand performance ceilings within structured educational datasets
NEST: Portable Seating and Positioning Device
Get PDFAbstract Seating and positioning are essential components of wheelchair technology that provide comfort, support, and protection for the user throughout all of their Activities of Daily Living (ADLs). These components are even more crucial for nonambulatory, full-time wheelchair users, and/or those with reduced-to-no sensation who are at a high risk of discomfort, injury, and even death from improper seating and positioning setups. However, there are several instances when this group of users must be out of their wheelchairs and seating and positioning setup, such as when traveling by car or plane, when participating in adaptive sports, and when engaging in social activities. Because of this oversight, users are at risk for pressure injury, sores, and even death when out of their chairs. This results in an impossible decision: choosing between pressure injury or facing isolation and depression from missing out on valuable life moments. NEST was designed in response to this problem and is a seating and positioning device that provides cushioning, protection, and support when outside of the wheelchair and is entirely customizable and portable. NEST incorporates an alternating pressure air cell cushion, backrest, sideguards, non-slip bottom, and several options for positioning accessories. NEST has two base models that inform the general use case: an active model and a stationary model. The active model features a rounded backrest to help facilitate movement when participating in activities like adaptive sports. The stationary model features a flat-top backrest to provide stability and additional support for the user when traveling. NEST results in a better quality of life for the user in terms of their social, physical, and mental health. The chance of acquiring pressure injury while participating in ADLs outside of their wheelchair would be reduced when using NEST, allowing users to fully participate in all of their desired activities with comfort, dignity, and safety
Towards Vision Intelligence-Based Liver Surgery
Get PDFDespite advancements in surgical interventions, modern procedures still heavily rely on surgeons\u27 expertise, requiring extensive training while maintaining limited accuracy. Although image guidance systems have been developed, “GPS-like” surgical navigation systems have yet to become standard practice due to their high costs and accuracy limitations. This thesis aims to enhance surgical navigation and mitigate some of its current limitations by leveraging vision intelligence: integrating image processing, modeling, and computer vision to extract rich, underlying information from images. The key to achieving these improvements lies in improving surgical perception, which involves understanding the spatial relationships between the endoscopic camera, surgical instruments, and surgical targets. This research contributes to four fundamental vision tasks for surgical navigation: depth perception, endoscope tracking, surgical instrument identification and segmentation, and registration of pre- and intraoperative data. Traditional depth perception methods struggle with featureless tissue surfaces, while learning-based approaches often require ground truth depth, which is difficult to obtain. Additionally, learning-based methods may lack robustness in cross-domain applications, and endoscopic images with calibrated camera parameters are rare. To address these challenges, we introduce an unsupervised optical flow-based depth estimation method for stereo endoscopes, eliminating the need for ground truth depth and camera calibration during training. Furthermore, a disparity framework is proposed, incorporating physical constraints and learning-based priors to enhance the accuracy of depth estimation in cross-domain settings. Endoscopic imaging features a limited field of view, making it difficult to track the camera pose and accurately align intraoperative point clouds from different perspectives. While hardware-based tracking systems provide potential solutions, they require additional instrumentation and add invasiveness. To overcome these limitations, this work proposes a hybrid framework that combines learning-based dense depth estimation with visual odometry, enabling precise endoscope tracking and surgical scene reconstruction. State-of-the-art methods for surgical instrument identification and segmentation depend on supervised learning with pixel-level dense annotations, which are labor-intensive to obtain. To reduce annotation dependency, in this thesis we explore a scribble-based, weakly supervised approach that serves as a precursor to more efficient and scalable surgical instrument segmentation. Lastly, existing image-guided surgery systems rely on manually performed rigid registration, which is both error-prone and time-consuming. To improve registration accuracy and efficiency, this work investigates the use of learning-based feature descriptors for automatic rigid registration. Beyond rigid registration, non-rigid registration is critical for correcting tissue deformations and ensuring accurate mapping of preoperative structures, such as tumors and vessels, to intraoperative scenes. To this end, we propose a biomechanical-model-based non-rigid registration method that offers a simplified formulation, simple hyperparameter choosing, and improved accuracy, all without requiring manual interaction. This research advances the integration of vision intelligence into surgical navigation and addresses key challenges in surgical perception, paving the way for more accurate, cost-effective, and accessible image-guided surgery systems
PM2.5 Forecasting at U.S. Embassies and Consulates Worldwide Using NASA Model Powered by Machine Learning
Get PDFAir quality forecasting is crucial for public health, especially in rural, suburban, and developing areas lacking reliable monitoring data. Hybrid monitoring (surface, satellite, and models) offers a scalable, cost‐ effective solution for tracking pollution and trends. This work presents a machine learning model that integrates ground measurements with global model outputs assimilating satellite observations to forecast air quality. Ground measurements of fine particulate matter (PM2.5) from over 60 U.S. embassies and consulates were used to calibrate global model outputs for local air quality forecasting. Multi‐channel input data was prepared using the Goddard Earth Observing System forward processing for meteorology and aerosol forecasts over 72 hr. An advanced convolutional neural network addressed high‐dimensional data and nonlinearities between inputs and outputs. A global model was developed and fine‐tuned with continent‐specific local models. The global model achieved Root Mean Squared Error (RMSE) and slope of 5.64 μg/m3 and 0.96, respectively. Local models showed improved performance with RMSE of 3.21 μg/m3 and slope of 0.98, outperforming the global model in Air Quality Index predictions by 6.57% in accuracy and greater stability during variability. The forecasts are publicly accessible via an application programming interface, providing global air quality predictions for 269 U. S. embassy and consulate sites to support public health and operational planning
Waveform Systematics in Moderate Mass Ratio Binary Black Hole Systems
No full textGravitational wave (GW) observations of binary black hole (BBH) mergers by advanced LIGO (aLIGO) have transformed our understanding of compact objects (COs). As detector sensitivity improves, new populations consisting of moderate to high mass ratio binaries will become increasingly accessible. Parameter estimation (PE), a process that extracts source properties such as masses and spins from GW signals, is critically dependent on accurate waveform models. However, current models may use approximations that introduce systematic errors for binaries with mass ratios between 0.05 and 0.5, affecting parameter accuracy and scientific interpretation. Using the Rapid Iterative FiTting (RIFT) algorithm, this work quantifies waveform systematics in this mass ratio regime and evaluates their impact on parameter inference. Addressing these challenges is essential for reliable interpretation of current and future GW observations, including those from next-generation detectors like LISA
EXPLORING THE RELATIONSHIP BETWEEN SOCIAL MEDIA AND ANXIETY IN DEAF COMMUNITY STUDENTS: A MIXED-METHODS APPROACH.
No full textThe mental health of Deaf individuals, particularly in relation to social media engagement, is an understudied area amid growing global concerns about anxiety disorders. Deaf individuals face unique challenges, including communication barriers, social isolation, and stigma, all of which can heighten anxiety levels. Although social media offers valuable opportunities for connection and self-expression, it may also amplify stress and feelings of inadequacy when accessibility barriers persist. This mixed-methods study examined anxiety among Deaf college students and explored the cultural validity of two widely used anxiety measures—the State-Trait Anxiety Inventory (STAI) and the Beck Anxiety Inventory (BAI). Quantitative data were collected through standardized anxiety surveys, and qualitative interviews explored students’ lived experiences with social media. Findings inform the development of culturally sensitive mental health assessments and highlight the need for more accessible and inclusive digital environments for Deaf user
CFD SIMULATION AND FIELD VALIDATION OF STACK EFFECT IN ELEVATOR SHAFTS AND THE EFFECTS OF AIRFLOW ON ELEVATOR DOOR OPERATION
No full textStack effect can lead to high pressure difference across elevator doors in high-rise buildings. This pressure difference leads to adverse effects during door operation and can lead to malfunctions. This research performs a comprehensive analysis of this phenomenon using both on site field measurements and computational fluid dynamics (CFD) simulations. Field data on pressure and temperature were collected from two high-rise buildings in Dubai, UAE, to establish baseline conditions and validate the CFD models. Multiple CFD models were developed to simulate the stack effect and its interaction with other phenomena in the shaft such as the piston effect. These were full-scale shaft models used to analyze the superposition of stack and piston effects, detailed car models to investigate airflow and pressurization around the elevator during car movement and a high-fidelity door model to examine forces during the opening and closing cycle. The findings confirm that the pressure induced by the piston effect is cumulative with the baseline stack pressure but dissipates entirely as the elevator decelerates and comes to a complete stop. It was discovered that the location of natural ventilation cutouts on the elevator car significantly impacts in-car pressure during movement. Furthermore, the simulations revealed that aerodynamic forces on the doors peak during the initial opening phase, a phenomenon which can be attributed to the Coanda effect caused by the landing door\u27s fire chicane and the design of the elevator lobby and the shaft. Based on these insights, several passive and semi-active design modifications were proposed and simulated. Two solutions were proven to be the most effective, first being enclosing the gap between the car and landing doors to create a pressure-equalizing air pocket, which reduced net forces by 5% passively and up to 30% when assisted by the car\u27s ventilation fan, and second, introducing vertical slits in the landing doors, which achieved up to an 8% reduction in force. This study provides a validated understanding of the airflow phenomena affecting elevator doors and offers tangible, data-driven design solutions for manufacturers to mitigate operational issues
Capturing Community Memory: A Juxtaposition of Oral History and Historic Building Preservation
No full textPreserving local history has always been a challenge for small communities. Traditional preservation methods like historic building preservation can be costly and take a lot of time for local historical societies. These historic sites are important to the culture of the local citizens and deserve to be saved. Using The Castle on the Hill in Dansville, New York, as a case study, oral histories will be captured to explore the role of oral history in preservation. The Castle on the Hill plays an important role in shaping Dansville’s identity as a town. In the past, the Castle has served as a hub for health and wellness in New York State. Now, the Castle stands in decay despite continued community support and precious revitalization efforts. Through three longform oral histories with individuals closely connected to the Castle, this research examines how personal narratives can shape public memory, inspire local advocacy, and provide insight into the cultural impact of lost or endangered historical sites. By analyzing themes of small-town identity, economic change, and generational shifts in historical consciousness, this study assesses whether oral histories can contribute to preservation efforts in meaningful ways. In doing so, it offers a framework for local historical societies and preservationists to consider oral history as a viable tool for safeguarding community heritage when physical restoration is not feasible
The Drifter: Creating a Family Archive to Rediscover Lost History
No full textWhen history goes undocumented or removed from public access, there is always a question as to how to preserve it. This thesis focuses on the creation of a family archive as a means to preserve and rediscover lost history. When lost history is recovered stories can be completed and we, as a civilization, can gain a better understanding of what our history is. This will be showcased through the creation of the Jerrold Smith Family Archive. Jerrold Smith (1941-1986) a native Rochesterian, grew up with a love of crafts and creating new things. When he passed, he left behind a collection of creations and correspondence that expressed his devotion to crafts. The Jerrold Smith Family Archive will encapsulate the life and career of silversmith and craftsman Jerrold Smith of Rochester, New York. Including notable achievements at Bausch & Lomb in their sunglasses division, Ray-Ban, and his ownership of Jerrold Smith Custom Designs, a local jewelry store. The creation of the Jerrold Smith Family Archive will be supported by documents examining the importance of family archives and preserving lost history. By combining the theory behind family archives with the practical application of the Jerrold Smith Family Archive I will evaluate the importance and act of preserving lost history