ScholarWorks (California State University)
Not a member yet
87523 research outputs found
Sort by
A Handbook for Educators and Guardians: Getting Students to School
No full textChronic absenteeism is a growing concern in high schools across the United States, particularly in the wake of the COVID-19 pandemic. Defined in California Education Code § 48263.6 as missing 10% or more of the academic year for any reason, chronic absenteeism is closely associated with lower academic performance, increased dropout risk, and long-term emotional and behavioral difficulties. Despite its wide-reaching effects, high school absenteeism remains under-addressed. Many schools and families lack access to cohesive, evidence-based tools that can intervene before patterns of disengagement become entrenched. This project introduces a comprehensive handbook for stakeholders, including educators, parents, mental health professionals, and school administrators, grounded in Bronfenbrenner's ecological systems theory and the Multi-Tiered System of Supports (MTSS) framework. By translating research into practice, the handbook provides real-world tools to address the multifaceted causes of absenteeism, such as student mental health, housing instability, academic disconnection, and limited transportation access. This resource guides systems-level collaboration through targeted, trauma-informed, and culturally responsive strategies to support consistent school attendance and promote equitable educational access for all students
AI-Based Resume Parser and Job Matcher
No full textThis graduate project develops an AI-based system to streamline and improve the recruitment process by automating resume parsing and candidate-job matching. Leveraging advanced natural language processing techniques, specifically transformer-based models such as Hugging Face Transformers, Sentence Transformers, and zero-shot classification, the system accurately extracts critical information like skills, education, and professional experiences from diverse resume formats. The extracted information is semantically matched with job descriptions to recommend suitable job opportunities for candidates and identify the best candidates for employers. Numerical experiments validate the system's high accuracy and efficiency, demonstrating significant improvements over traditional manual screening and keyword-based methods. Key metrics, including precision, recall, and F1-score, confirm robust performance, particularly in skill extraction and job matching tasks. The AI-driven approach not only accelerates recruitment timelines but also enhances fairness by minimizing human bias. However, the system faces challenges in parsing complex or unconventional resumes and is primarily optimized for English-language content. Future enhancements recommended include expanding multilingual capabilities, broadening training datasets, and integrating continuous feedback mechanisms. Overall, this project highlights the transformative potential of AI in recruitment, paving the way for more efficient, accurate, and inclusive hiring processes
Machine learning on digital wearable data in congenital heart disease patients - Analyzing the use of digital wearable data on ventricular health of CHD patients
No full textWearable devices, such as fitness trackers, provide continuous monitoring of physical activity and insights into health disparities. This study analyzes Fitbit data from individuals with congenital heart disease using the All of Us Research Program [1]. Cohorts were selected based on electronic health record and Fitbit data availability (2015-2019), excluding pandemic-related distortions. The final sample included 537 CHD and 36071 no-CHD participants. Data preprocessing was conducted to ensure consistency and reliability before statistical analysis. ANOVA results demonstrate that CHD participants had significantly lower daily step counts than those without CHD (p = 0.013240), making step count the most reliable metric for detecting CHD-related activity limitations. However, sex at birth, age, BMI, and geographic location showed the greatest overall impact on activity levels. CHD status was only significant for step count and lightly active minutes, with no observable difference in more intense or sedentary activity. These findings highlight daily step count as a key distinguishing feature for CHD status and emphasize its potential for machine learning applications, particularly in k-clustering methods [29], early detection and remote monitoring of CHD patients using wearable devices
Map Bot: Joystick-Controlled LiDAR SLAM Explorer
No full textThis project mainly focuses on building a differential drive robot to create a map and detect objects as it maps and localizes the environment using a LiDAR sensor. The robot is run using a joystick. The velocity commands from a joystick are processed through ROS 2 nodes and transmitted through serial communication to a microcontroller that controls motor actuation through an H-bridge driver. The project is run in simulation and real life; the output is viewed in RViz and Gazebo to visualize the behavior and environment. Raspberry Pi 4 is used as it's more efficient, with ROS 2 Humble, providing a 90% accuracy. Arduino UNO is used for motor control. A joystick is used to run the robot, and the output is viewed in RViz2 utilizing the concept of the SLAM algorithm, and the robot's movements in terms of points are plotted in the map
Stock Price Forecasting Using LSTM and GRU and Sentimental Analysis
No full textThe dynamic nature of the financial markets and the increasing volume of huge data which is generated have posed some challenges to stock price prediction. This research work accords with the challenges of stock price prediction using news articles and social media textual sentiments as well as univariate time series using Long Short-Term Memory and Gated Recurrent Units (Singh, G., 2022) (GRU) models. The historical prices are integrated with micrometeorological indicators that reflect the mood of the market, which should lead to higher forecast accuracy. Namely, Google Colab was applied to develop an end to-end system with a simulated sentiment dataset attached to historical stock data collected from public APIs. Standard metrics were used for the evaluation of the models, TensorFlow and Keras frameworks were used for building the models and the models were trained for a long time. In this report, the author presents the methodological approach, the architecture of the system, and the experiments that were conducted. Thus, it is conformed that utilising the hybrid model can enhance the understanding of temporal dependencies and market sentiment to generate better investment decisions for investors, financial analysts, and strategists. Furthermore, it has been suggested to utilize hybrid models that combine Convolutional Neural Networks (CNNs) with Gated Recurrent Units (GRU) and Long Short-Term Memory (LSTM) in order to capitalize on the advantages of each architecture. In time series data, CNNs are good at capturing short-term dependencies and pertinent characteristics, but Gated Recurrent Units (GRUs) and Long Short-Term Memory (LSTMs) are good at managing sequential data (Singh, G., 2022) . Research has demonstrated that this combination enhances the ability to forecast changes in stock prices
Analytical Modeling for SOI LO MESFET for High Voltage and RF Power Applications
No full textThis document provides a comprehensive analytical modeling and simulation investigation of Silicon-On-Insulator (SOI) Metal-Semiconductor Field-Effect Transistors (MESFETs), focusing on the impact of Local Oxide region engineering on improved high-frequency and high-power device performance. This study carefully examines the impact of LO layer positioning on essential electrical parameters, such as transconductance, drain-source resistance, gate-source capacitance, and drain current. A comprehensive analytical framework is developed, supported by MATLAB-based simulations, to derive and validate the relationships between device structure, material parameters, and electrical behavior. The incorporation of the LO region markedly alters the electric field distribution within the channel, leading to increased transconductance, diminished channel resistance, adjustable gate capacitance, and enhanced output current characteristics.The comparative analysis between conventional and LO-engineered SOI MESFETs demonstrates that optimal LO placement, particularly at the channel center, yields the most substantial performance improvements, as confirmed by both simulation and alignment with recent literature. Beyond characterizing the electrical properties, this work highlights the innovative role of field engineering in semiconductor device optimization. By carefully analyzing how the LO region alters internal electric fields and carrier transport, the thesis identifies clear design strategies to overcome longstanding trade-offs between power efficiency, voltage robustness and speed in SOI MESFETs. The results presented in this study provide actionable design guidelines for the development of high-efficiency, high-frequency SOI MESFETs suitable for RF amplification, power switching, and advanced integrated circuit applications
Mitigation of Herbivory Effects on Native Oak Seedling Survival in a Zoological Setting
No full textCaptive exotic species housed by zoos in native landscapes can create grazing pressures on native ecosystems, impacting seedling recruitment in foundation tree species. The loss of foundation species can have trickle-down effects on the entire ecosystem and, with slow-growing species such as oak trees, reduced recruitment can lead to generational effects. We conducted an exclosure experiment in a large safari park in Sonoma County, California, from 2023 to 2025 to assess impacts of active mitigation of small and large exotic herbivores on first-year survivorship and growth of blue oak (Quercus douglasii) seedlings. We studied the effects of herbivore treatment, greenhouse headstarting of seedlings, parent tree, canopy cover, and soil compaction on acorn survivorship and growth. To do so, we built twenty exclosures, each with twelve greenhouse-grown seedlings and twelve germinated acorns planted on site, with varying levels of protection from herbivory, canopy cover, and soil compaction. We found that herbivory exclusion had a significant impact on survivorship, with 0% of seedlings exposed to both small and large herbivores surviving, and 16.3% survivorship in seedlings protected from both small and large herbivores. Mitigating herbivory of small herbivores had a significant impact on basal stem diameter, resulting in first-year seedlings with approximately 1 mm larger basal stem diameter than seedlings not protected from small mammals. Greenhouse-grown seedlings had significantly higher survivorship than acorns planted on site. In addition, parent tree significantly affected survivorship, highlighting the importance of genetics in sourcing acorns for planting. These results demonstrate that basic mitigation efforts can significantly improve oak recruitment in the early seedling stage and effectively support efforts of zoological institutions towards conservation and education goals that include local sustainability of native species
From admission to graduation a guide for diversifying school psychology graduate programs
No full textThe field of school psychology has long struggled with a lack of diversity, with racially, ethnically, and linguistically (REL) diverse professionals being underrepresented in not only graduate programs but also the profession as a whole. This thesis project explores research-based strategies to diversify the field of school psychology by understanding the challenges faced by REL diverse graduate students in the broader field of education and developing a guidebook for school psychology graduate programs to implement multi-tiered systems of support (MTSS) framework. Furthermore, the guidebook is intended to give school psychology a starting point to increase the recruitment and retention of diverse students and support them from admission through graduation. Using the MTSS approach, the guidebook caters to targeted interventions at a universal, targeted, and intensive level to ensure diverse students have equitable access to school psychology graduate programs. Through providing research-based best practices, this guidebook strives to provide actionable recommendations and resources for school psychology graduate programs to facilitate the diversification of school psychologists that serve increasing numbers of diverse students in education
Expression and purification of cyp6a20, a cytochrome P450 protein associated with aggression modulation in D. Melanogaster
No full textEveryone is in constant search for the next anti-aging solution, from skincare to supplements, but could social behavior itself influence the rate of aging? Studies in humans and animals suggest that positive social interactions correlate with longevity, while negative social environments are linked to accelerated aging. In Drosophila melanogaster, aggression is a well-studied behavior regulated by genetic and environmental factors. One gene of interest, Cyp6a20, has been identified as a modulator of aggression. This gene encodes a cytochrome P450 enzyme, but its functional role remains unexplored. This study aimed to express and purify Cyp6a20 protein, laying the groundwork for future biochemical and structural analyses. Our findings show that Cyp6a20 knockout (KO) flies (loss of Cyp6a20 function) displayed increased aggression along with sex-specific changes in pheromones. These behavioral and pheromone changes were associated with reduced lifespan, suggesting a role for Cyp6a20 in linking social behavior and aging. Iterative optimizations to the expression and purification method led to the development of an effective Cyp6a20 purification protocol. Construct enhancements, including the addition of a linker sequence, significantly improved purification results. Additionally increasing the concentration of 5-aminolevulinic acid (5-ALA) during expression ensured heme incorporation and resulted in better Cyp6a20 expression. These optimizations provide a reliable expression and purification method, laying the foundation for future studies of the enzyme's biochemical activity and its role in pheromone signaling and aging
Improving Reading Comprehension Through Targeted Small Group Instruction for 3rd Grade
No full textThis project began with a concern I see every day: many elementary students, especially those in high-poverty schools, are not getting the necessary support with reading comprehension. Teachers often have materials for phonics and decoding, but there are fewer practical tools for helping students actually make sense of what they read. This project set aimed to minimize or fill that gap by creating a small-group guide that teachers can easily use. The guide includes lessons on seven key comprehension strategies: predicting, questioning, summarizing, clarifying, monitoring, inferring, and evaluating. Each lesson is structured around a simple framework, incorporating modeling, guided practice, independent work, and reflection. Each lesson also comes with tools like graphic organizers, rubrics, and checklists. The lessons were designed with Title I classrooms in mind, with scaffolds for English learners, supports for struggling readers, and texts that reflect students' cultures and experiences. As I worked on this project, I kept coming back to the idea that literacy is about more than just academics. It is about equity, identity, and access. This guide not only helps teachers feel more confident about teaching comprehension but also helps students see themselves as capable readers. Ultimately, the goal is for students to read with understanding, connect with texts, and carry those skills into their future learning and lives