About Me
My name is Clement Ampong, and I hold a Master’s degree in Mathematics with a concentration in Scientific Computing from Georgia State University. My academic and professional journey is fueled by a passion for problem solving and a desire to use mathematics to address real world challenges. I have experience in teaching, research, and collaborative projects, including forecasting disease dynamics, modeling biological growth, and developing numerical methods. These experiences have strengthened my analytical thinking, programming skills, and ability to translate challenging mathematical ideas into practical solutions.
I am highly organized, self motivated, and committed to continuous learning. My long term goal is to pursue a Ph.D in Applied and Computational Mathematics, where I aim to contribute to innovation in fields like healthcare, technology, and education. Beyond my academic pursuits, I find joy in mentoring students and helping others develop confidence in mathematics. I believe in creating collaborative, supportive learning environments that encourage curiosity, resilience, and growth. In my free time, I enjoy watching soccer, exploring new places, and watching movies, which help me relax and stay inspired.
Education
M.S., Mathemathics
Georgia State University (August 2023 - May 2025)
Advisor: Alexandra Smirova
B.S., Mathematics
University of Cape Coast (August 2018 - November 2022)
Advisor: Henry Amankwah
Research Interests
- Machine Learning and Data Science
- Numerical Methods and Scientific Computing
- Epidemiological Modeling
- Optimization and Applied Mathematics
- Deep Learning
Projects
Forecasting Growth Trajectories Using Phenomenological Growth Models
Applied the QuantDiffForecast toolbox to model and forecast time-series data of daily monkeypox cases in the USA using phenomenological growth models. This project aimed to accurately predict growth trajectories, understand the dynamics of disease spread, and assess the practical implications of the models in guiding public health decisions. The work demonstrates the effectiveness of advanced mathematical tools in addressing emerging health challenges and provides a framework for applying similar methodologies to other infectious diseases.
View DetailsParameter Estimation and Forecasting Using ODE Models
Performed parameter estimation and forecasting for the 1918 influenza pandemic using the SEIR model and QuantDiffForecast MATLAB toolbox. Nonlinear Least Squares (NLS) and Maximum Likelihood Estimation (MLE) were employed to estimate the transmission rate and generate forecasts with quantified uncertainty. The study evaluated the performance of both methods, demonstrating that MLE provided tighter uncertainty bounds and superior reliability. This work underscores the importance of advanced mathematical modeling techniques in epidemic forecasting and their potential to enhance public health planning.
View DetailsSelection in a Population of Three Phenotypes
Analyzed the effects of selection in a population with three distinct phenotypes (white, gray, and black-winged moths) using a recurrence relation derived from a Punnett square. The study examined the dynamics of allele frequencies under varying selection pressures and identified conditions for phenotype coexistence or extinction. Cobweb diagrams were utilized to visualize long-term allele frequency behaviors, revealing that gray-winged moths survive only under specific selective advantages. This work highlights the application of mathematical modeling in understanding population genetics and the impact of selection on genetic diversity.
View DetailsData-Enabled Modeling of an Epidemic: Smallpox and COVID-19
Analyzed the dynamics of the smallpox epidemic in Abakaliki, Nigeria (1967), and the COVID-19 pandemic in the United States (2020) using the SIR model. Key parameters, such as transmission and recovery rates, were estimated to understand disease spread and predict intervention strategies. For smallpox, vaccinating a minimum of 15 individuals would have reduced the susceptible population below the epidemic threshold, effectively halting the outbreak. For COVID-19, the analysis determined that vaccinating or quarantining over 11 million individuals could have stopped the pandemic. This project underscores the importance of vaccination and highlights the limitations of deterministic models in capturing human behavior and uncertainty.
View DetailsPredicting Malignancy in Breast Cancer Using Gaussian Process Regression
Conducted an in-depth analysis of clinical features to predict breast cancer malignancy using Gaussian Process Regression (GPR). The project captured nonlinear relationships between features and improved diagnostic accuracy. Key predictors, including mean radius, mean area, and mean perimeter, were identified, enhancing the interpretability of the model through robust prediction intervals. The results demonstrated the effectiveness of GPR in medical prediction tasks, highlighting its potential for applications in cancer diagnosis and broader healthcare analytics.
View DetailsEffectiveness of Chelation Treatment with Succimer on Blood Lead Levels in Children
Developed a robust SAS model to analyze longitudinal data on blood lead levels in children, focusing on assessing the effectiveness of Succimer compared to a placebo. This study aimed to evaluate the treatment's impact over time, identify patterns in lead reduction, and provide statistical insights into its efficacy and safety. The project highlights the importance of data-driven approaches in public health interventions and offers valuable recommendations for future clinical applications.
View DetailsDirect and Iterative Methods to Solve Two-Point Boundary Value Problems
Designed and implemented a MATLAB model to solve two-point boundary value problems using the conjugate gradient method. The project aimed to compare the performance, accuracy, and computational efficiency of the method when applied as a direct solver versus an iterative approach. This work provides valuable insights into the practical applications of numerical methods in solving complex differential equations and highlights the trade-offs between solution techniques in mathematical modeling.
View DetailsForecasting Influenza Dynamics in the US
This study explored advanced forecasting techniques for influenza outbreaks across the United States using the n-sub epidemic model and the STAT MOD PREDICT toolbox. Weekly influenza case data from May to October 2024, collected through CDC surveillance and wastewater monitoring, were modeled to assess outbreak dynamics. Predictive models, including ARIMA, GLM, GAM, Prophet, and ensemble approaches, were employed to generate region-specific forecasts. The analysis identified ARIMA as the most effective model for forecasting influenza dynamics across all regions, achieving the lowest error metrics (MAE and MSE) and reliable 95% prediction intervals. Weighted ensemble models also demonstrated strong performance in areas with complex epidemic patterns. By leveraging advanced statistical methods, the study provided accurate short-term forecasts to support early intervention and resource allocation strategies. This work underscores the critical role of robust predictive modeling and surveillance data integration in guiding public health responses. The findings emphasize the potential of these methodologies to mitigate the societal impact of influenza outbreaks by enabling timely and informed decision-making.
View DetailsApplication of Linear Programming in Profit Maximization of Asanduff Construction
Applied linear programming techniques to optimize profit for Asanduff Construction, a block manufacturing company producing four types of blocks (8-inch, 6-inch, 5-inch, and 4-inch). By incorporating data on production costs, labor, time, and sales into a linear programming model, the study employed Excel Solver and the simplex method to identify the optimal production strategy. The analysis revealed that prioritizing the production of 8-inch, 6-inch, and 5-inch blocks while reducing resources allocated to 4-inch blocks maximized daily profits to GHC 5564. This optimization resulted in a 27.1% increase in daily profits, showcasing the effectiveness of linear programming in driving significant financial improvements. Additionally, shadow price analysis provided insights into constraints impacting profitability, guiding better resource allocation. This work highlights the value of mathematical optimization in enhancing profitability, improving operational efficiency, and supporting strategic decision-making in manufacturing processes.
View DetailsPublications
Statistical and machine learning approaches in U.S. public health surveillance: An empirical review
Public health surveillance in the United States depends on accurate and timely methods for detecting disease outbreaks and protecting population health. Traditional statistical approaches such as regression analysis, Bayesian modeling and time series analysis have served this purpose for decades, but they struggle to process the large, complex and unstructured datasets that modern public health challenges produce. This review paper examines how statistical and machine learning methods are applied in U.S. public health surveillance and explores their potential to improve health equity across underserved populations. A systematic review of peer-reviewed literature published was conducted across PubMed, Web of Science, Scopus and Google Scholar using PRISMA guidelines, with independent reviewers screening and assessing all included studies for methodological quality. The findings show that supervised learning algorithms, unsupervised clustering methods and deep learning architectures consistently outperform traditional methods in outbreak detection, disease forecasting and anomaly identification. AI-augmented models achieve detection accuracy scores of 1.000 and reduce security breaches by up to 75 percent compared to conventional approaches. Explainable AI tools further improve transparency and support equitable deployment by identifying and mitigating algorithmic bias. However, data interoperability gaps, inadequate workforce training and unresolved ethical concerns continue to limit widespread implementation. This review concludes that sustained investment in digital infrastructure, interdisciplinary collaboration and robust regulatory frameworks are essential for integrating these technologies into public health practice in a manner that is effective, equitable and accountable.
View PublicationForecasting Influenza trends in the United States using Public Surveillance Data and Applied Mathematical Models
Seasonal influenza imposes a substantial public health and economic burden in the United States. The CDC estimates between 9 million and 45 million illnesses, 140,000 to 810,000 hospitalizations and 12,000 to 61,000 deaths annually, depending on epidemic severity. Notwithstanding comprehensive surveillance through the CDC's Influenza-Like Illness Network (ILINet), which encompasses over 3,000 healthcare providers across all 50 states, accurate prediction of epidemic timing and intensity remains challenging. This unpredictability complicates resource allocation and preparedness efforts. This study analyzed weekly Weighted Influenza-Like Illness (WILI) data from the 2010-2020 influenza seasons to develop reliable short-term forecasting models. Three established time-series approaches were compared: Seasonal Autoregressive Integrated Moving Average (SARIMA), state-space models, and Holt-Winters exponential smoothing. Performance was evaluated using mean absolute error (MAE), root mean squared error (RMSE) and mean absolute percentage error (MAPE) across one-week to four-week forecast horizons. Temporal analysis confirmed characteristic seasonal patterns with peak activity typically occurring between December and February and minimal activity during summer months. SARIMA models achieved the lowest forecasting errors and accurately predicted peak timing within one week for 85% of seasons. State-space and Holt-Winters methods showed reduced accuracy during epidemic surges despite adequate performance during baseline periods. These findings demonstrate that SARIMA-based forecasting systems provide actionable advance predictions to support hospital surge planning, antiviral stockpiling, targeted vaccination campaigns and public health messaging, which potentially reduce morbidity and mortality through enhanced preparedness.
View PublicationContact
Email: campong1@gsu.edu
Address: Georgia State University, Atlanta, GA 30328, USA
Phone: (240) 840 0372