Michael Asante Ofosu

ABOUT ME

Michael is a research and teaching assistant at the Department of Statistics and Actuarial Science at Kwame Nkrumah University of Science and Technology specializing in Multivariate Data Analysis, Bayesian Analysis, Machine Learning, Statistical Computing, Biostatistics, Probability and Statistics, Linear Models, and Group Theory

His research interest focuses on Bayesian Statistics, Multivariate, and Applied Statistical methods for modeling real-world problems such as Health, Transportation, and Climate Change. He is active in multidisciplinary research in Environmental, Natural Science, and Public Health.

Technical Proficiency

Statistical programming: Proficient in R, Python, SPSS, STATA, and MATLAB for advanced data analysis and modeling
Data visualization: Skilled in Power BI and MATLAB for impactful data visualization and dashboards
Machine learning: Experienced with supervised and unsupervised techniques including regression, random forests, neural networks, clustering, PCA
Structural equation modeling: Extensive experience building and analyzing SEMs using SmartPLS, AMOS, and lavaan.
Document preparation: Highly proficient in LaTeX for typesetting research papers, reports, and presentations.
Microsoft Office: Advanced skills in Excel, Access, Word, and PowerPoint, including VBA.

Additional Skills

Video editing: Proficient in Filmora and Wave Pad for video production and audio editing
Graphic design: Experienced with Adobe Photoshop, Illustrator, After Effects, PixelLab, and Canva for digital design
Research: Proven ability to interpret data and derive meaningful insights from statistical analysis
Project planning: Skilled at setting goals, developing timelines, and managing projects effectively

Research Experience

Undergraduate Thesis: Statistical Modeling of Students' Energy-Saving Behavior

Employed advanced techniques such as Structural Equation Modeling and Machine Learning.
Analyzed and modeled students' energy-saving behaviors using R and STATA.
Developed a comprehensive model, contributing to more effective energy-saving programs and sustainability initiatives within educational institutions.

Measles Vaccination Coverage Cluster Analysis

Performed hierarchical clustering analysis in R on district-level data including vaccination rates, demographics, healthcare access, and utilization.
Identified distinct clusters and profiles of districts with varying measles vaccination coverage. Provided context and implications for the clustering patterns based on healthcare infrastructure and population characteristics.

Understanding Postpartum Exercise and Body Image Satisfaction

Conducted regression analysis and ANOVA tests in STATA to model associations between age, race, education level, and postpartum exercise frequency/duration.
Key findings showed postpartum exercise for 2+ hours per week was linked to higher body image satisfaction across all demographics. This highlights the importance of physical guidance and community programs to encourage postpartum physical activity.

Interdisciplinary Research Collaborations and Analyses: Brent Crude Oil Price Forecasting

Developed time series forecasting models including exponential smoothing and ARIMA models in R to predict monthly Brent crude oil prices.
Findings showed the ARIMA model performed better based on lower RMSE. The analysis provides important insights for stakeholders across the oil and gas value chain.

Unsupervised Machine Learning Analysis of Bone Mineral Content

Utilized K-means clustering, hierarchical clustering, and principal component analysis in R to classify archaeological bone specimens based on calcium/phosphorus ratios.
Identified distinctive clusters representing dietary patterns and predisposition for bone diseases among ancestral groups.

Exploring Specimen Data Relationships

Collaborated with archaeology students to analyze bone specimen data from ancestral sites across Ghana.
Performed correlation and regression analysis in R to assess relationships between bone calcium/phosphorus levels, presumed dietary patterns, geographic location, and disease prevalence.
Key findings revealed associations between high cereals diets in certain regions and elevated phosphorus levels, indicating malnutrition and rickets prevalence

Predictive Analysis for Forecasting Stock Returns

Assisted in time series predictive modeling using MtNEGH open-source financial data to forecast monthly returns for stocks listed on the Ghana Stock Exchange.
Developed ARIMA and LSTM models in Python, optimizing hyperparameters based on the lowest RMSE.
The deep learning LSTM model performed better, capturing complex nonlinear patterns in the time series data.

Impact of Chatbot Conversational Agents on University Student Engagement

Collaborated on a study to assess the impact of chatbot conversational agents on student engagement at KNUST.
Analyzed survey data on student usage of and attitudes towards chatbots using structural equation modeling tools like SmartPLS and AMOS.
Key findings showed chatbot interactions increased active learning time but did not improve perceived engagement levels.

Associations of Anthropometric Metrics with Birth Weight and Placenta Histomorphology

Co-authored research using machine learning algorithms like random forests and Bayesian hierarchical modeling in R.
Analyzed associations between maternal anthropometric measurements during pregnancy and neonatal outcomes including birth weight and placental structure.
Key findings revealed certain mid-upper arm circumference thresholds during pregnancy associated with low birth weight babies.

Classification Models Assessing Patient Attributes for Borrower Default Levels

Contributed to research employing random forests and logistic regression modeling in Python to predict borrower default risk.
Analyzed patient health records data including demographics, diagnosis codes, and healthcare utilization metrics as predictors of loan default.
Key drivers of default identified were age, income level, and frequency of hospitalizations.