Mason Journals (George Mason Univ.)
Not a member yet
3256 research outputs found
Sort by
Synthesis and Structural Analysis of Bi2WO6 Catalysts under Different Conditions
No full textBi2WO6 is an attractive, multifunctional photocatalyst with significant applications in water splitting and the decomposition of organic contaminants. In this research, we aim to investigate the synthetic conditions for preparing Bi2WO6 polycrystalline powder samples by a solid-state reaction method. Balanced stochiometric quantities of high-purity chemicals of Bi2O3 and WO2 were mixed and ground together, pressed into a pellet, and then heated at high temperatures with a variety of heating profiles. This method was then repeated with WO2 replaced by WO3. Powder X-ray diffraction technique was used to determine the respective purities and impurities to determine which heating profile is best suitable
Optimizing Solar Microgrid Locations in Morocco Using Geospatial Analysis and Random Forest Machine Learning Techniques
No full textExpanding access to renewable energy in rural regions is critical for sustainable green development and equitable access to electricity. Morocco, one of Africa’s leading countries in renewable energy development, has invested in large-scale electric grids/projects such as the Noor Ouarzazate Solar Complex and national energy plans targeting 52% renewable electricity by 2030 (El Hafdaoui et al., 2025). However, rural Moroccan communities remain disadvantaged and dependent on fossil fuels. This study aims to address this gap by combining geospatial data analysis with machine learning in order to identify the optimal regions for solar microgrid development. Five key monthly datasets were obtained: net downward shortwave solar radiation (FLDAS) and cloud fraction/cover from NASA’s Moderate Resolution Imaging Spectroradiometer (MODIS) data on NASA Giovanni, as well as terrain elevation (SRTM), land cover classification (MODIS), and night-time light intensity (VIIRS) from Google Earth Engine. Each dataset was processed, normalized, and visualized using Python in Google Colab. A weighted composite suitability map was generated using matplotlib, emphasizing solar irradiance and terrain (weighted at 0.35 each), along with cloud cover and urban proximity via night lights (0.15 each), and land cover as a binary suitability mask. 100 labeled data points (50 suitable, 50 unsuitable) were manually selected using QGIS. A Random Forest ML classifier was trained and tested on these layers to predict suitability across the map of Morocco and yielded a classification map consistent with the composite map. This study demonstrates the potential of a data-driven approach for solar microgrid siting to support effective planning and equitable energy access in rural regions and serves as a foundation that can be further refined and adapted for broader applications
Drought Impact on Maize Yields in South Africa Mapped via Google Earth Engine and Python-Based Remote Sensing
No full textClimate change has highlighted global trends in droughts that directly affect rain‑fed agricultural foodavailability. Maize, a core crop in South Africa, plays a significant role in regional food value chains and ishighly sensitive to rainfall variability. Despite the importance of this relationship, no spatially explicit timeseriesanalysis of drought indicators and maize yield outcomes has yet been documented. This gap wasaddressed by estimating the impact of drought on maize yield in South Africa through remote sensing dataanalyzed within Google Earth Engine (GEE) using Python. NDVI from MODIS, rainfall from CHIRPS, and landsurface temperature served as the primary variables. Time series data spanning 2015–2018—encompassingboth El Niño and non‑El Niño years—for key maize-producing provinces such as Mpumalanga and the FreeState were extracted. These datasets were merged with FAO and Stats SA maize yield data and examined usingcorrelation analysis, linear regression, and anomaly detection methods. A clear negative correlation emergedbetween temperature and drought severity with NDVI during El Niño seasons, accompanied by statisticallysignificant yield losses. Visualization via Geemap and folium revealed spatial patterns of drought intensity andyield reductions. These findings underscore the potential of open-source tools to support policy developmentand climate-resilient agricultural planning in vulnerable communities
SEIR-GA: A Dynamic Model for Unmasking Gerrymandering’s Impact on Voter Turnout and Representation
No full textThis project introduces the SEIR-GA model (Susceptible, Exposed, Infected, Recovered, Gerrymandering, Activism), a new mathematical framework exploring how gerrymandered electoral districts affect voter turnout and democratic representation over time. Inspired by epidemiological models, SEIR-GA adapts these concepts to reflect the social and political processes that influence civic participation. The intercompartmental flow is governed by a coupled system of Ordinary Differential Equations (ODEs). In the model, individuals transition through various stages based on their awareness and response to gerrymandering. They begin in a "Susceptible" state, unaware of district manipulation. Upon becoming informed through media or other communication channels, they move into the "Exposed" state. Prolonged exposure without change may lead to the "Infected" state, in which individuals feel disillusioned and conclude their vote has little impact, resulting in disengagement. Some eventually reach the "Recovered" state, regaining civic motivation through education, activism, or community efforts that restore confidence in the process. A central feature of SEIR-GA is its ability to represent the conflict between political mechanisms that maintain power through unfair districting and citizen reform movements. The model captures how these forces interact over time, influencing voter turnout and political participation. SEIR-GA allows for the simulation of various political environments, from heavily gerrymandered regions to those with strong democratic safeguards. By quantifying both the suppressive and mobilizing effects of gerrymandering, this study provides a predictive tool for policymakers and advocates. Ultimately, it aims to strengthen voter representation and support democracy, advancing UN SDG Target 16.7 for inclusive, participatory, and representative decision-making
Analyzing the Effectiveness of AI-Generated Tags for Traditional vs. Simplified Chinese Users on Weee
No full textOnline grocery platforms are increasingly using AI-generated “popular” tags on their products like “Top ReorderedDimsum” to boost product engagement. However, little is known about how these tags perform across different linguisticand cultural user groups. This study investigates why such tags seem to drive higher sales from Simplified Chinese usersthan Traditional Chinese users on Weee, an online Asian supermarket. We focused on two tags: a scenario-based one(“MVP Seafood for Hotpot”) and a category-based one (“Top Reordered Dimsum”). We extracted reviews for the top fiveproducts under each tag using Weee’s internal review API, then cleaned and separated the data by language. Using Jieba,a Chinese text segmentation library, we identified the top 20 keywords in reviews for each group and compared theirdistributions. We found no significant differences in keyword patterns between Traditional and Simplified Chinese reviews,suggesting the tags are likely relevant to both groups. However, there were consistently far fewer Traditional Chinesereviews, which may limit the visibility or effectiveness of the tags for that audience. These findings point to data sparsity,in addition to content misalignment, as a possible reason for the differing sales impact of AI-generated tags acrosslanguage groups
Local Government Approaches to EV Transition: A Comparative Study of Planning, Funding, and Infrastructure in Five U.S. Jurisdictions
No full textThis study compares electric vehicle transition strategies in five local governments: Baltimore, Raleigh, Fairfax County, Rockville, and Prince William County. While all jurisdictions aim to electrify their public fleets and expand charging infrastructure, they differ in policy scope, funding sources, deployment strategies, and data availability. Raleigh has transitioned over 80 percent of its city fleet to clean fuels and uses solar mobile chargers in public parks. Fairfax County, with a suburban layout and dispersed demand, has committed to full fleet electrification by 2035 and installed over 100 Level 2 chargers at government sites. Baltimore passed a city law requiring a fully electric light duty fleet by 2030 and partnered with BGE for charger installation at no cost to the city. Rockville coordinates closely with Montgomery County and relies on microgrid infrastructure, scenario modeling, and energy as a service financing. Prince William County focuses on phased deployment tied to vehicle replacement cycles and has installed 10 Level 2 chargers ahead of major fleet expansion. All five regions use state and federal grants, equity-based site selection, and permitting reforms to support deployment. However, only some have conducted detailed cost analyses or updated standard operating procedures to reflect the operational demands of electrification. These findings show that while foundational strategies are often shared, localized factors such as fleet size, geography, existing infrastructure, and institutional capacity shape how EV transition plans are developed and implemented
Application of Multi-Layer Perceptron (MLP) Machine Learning Model in Determining Mental Health Illness Severity
No full textSpikes in mental health (MH) awareness have led to the development of the 988 hotline, which was created to personally assist with MH issues. Hotline operators analyze a patient's specific MH crisis, illnesses, and other provided information, all to make a fast triage judgment on the severity of the patient's condition. Determining condition severity requires holistic consideration of factors (socioeconomic, specific illnesses, etc.) that one cannot instantaneously process; however, machine learning (ML) models can expedite the analysis. This research explores Keras’s Sequential multi-layer perceptron neural network model’s use in determining severe mental illness (SMI) in patients (i.e., illness that substantially interferes with a person's life and function). Using client-level data from SAMHSA's (Substance Abuse and Mental Health Services Administration) 2023 dataset, the model was trained on over 50,000 adult patients to predict the SMI target field (serious or not serious). Using socio-economic features (age, race, and gender) and known MH conditions of patients, the model accuracy is 85.44% in concluding the presence of SMI in a patient, with 92.53% precision and 85.50% recall in determining SMI patients and 73.47% precision and 85.32% recall in determining non-SMI patients. This study also considers factor importance when determining a patient's SMI, removing certain indicators to determine the impact on model performance. For instance, the removal of the patient's MH1 (first diagnosed MH condition) in consideration tanks model accuracy down to 78% with 89.01% precision and 76.42% recall in determining SMI patients and 61.47% precision and 79.94% recall in determining non-SMI patients. While findings don’t support artificial intelligence (AI) full replacement of 988 operators, they show that AI models can have value as tools for operators in assisting an accurate assessment of a patient’s mental state efficiently. For future improvement of the ML model, datasets with anecdotal behavior descriptions will be used in training for increased model accuracy
Comparative Analysis of LSTM and XGBoost ML Models for Short-term Rainfall Forecasting
No full textFloods are the most widespread and frequent of all weather-related natural disasters. Accurate short-term rainfall forecasting is essential for flood management and disaster preparedness. Traditional models often struggle to capture the variability and nonlinear patterns of rainfall, and they can be computationally intensive. Machine learning offers a promising alternative due to its ability to learn temporal dependencies in time series data. This study investigates deep learning and tree-based approaches to rainfall prediction by comparing LSTM neural networks and XGBoost models, along with a combined ensemble model. Using data from the IAD Airport station in Houston, Texas, models were trained to predict the next hour’s rainfall using the previous 24 hours. The dataset used spans from 2012 to the present, with roughly the first 10 years used for training and the rest for evaluation. Evaluation was based on MAE, RMSE, and R². Results show that while all three models produced similar accuracy, with over 96% of predictions within ±1 mm of actual, the LSTM model achieved MAE/RMSE of 0.2287/1.5389 mm; XGBoost scored 0.2174/1.4622 mm; the ensemble outperformed both with 0.1944/1.3761 mm and an R² of 0.5222, an error reduction of about 12–15%. These findings suggest that ensemble learning improves short-term rainfall prediction and offers a promising real-time rainfall forecasting method. Future work could explore incorporating additional environmental variables such as humidity or wind to further enhance predictive performance under diverse meteorological conditions