John Njuguna

I’m an Analytical Data Scientist with 1 years of technical Data Scientist Apprentice (Part-time, weekends) experience through ALX Africa. I focus on Statistics and Programming, turning large datasets into clear, decision-ready insights through machine learning and data visualization.

In my work, I’ve built and benchmarked predictive models using SQL, Python, and Power BI—evaluating Logistic Regression, Random Forest, Decision Tree, and KNN using precision, recall, F1-score, and ROC AUC. I also remedied severe class imbalance (3.4% failure rate) with RandomUnderSampler, improving minority-class failure detection and helping models perform on real-world problems (reported 0.62% f1 score) using a stratified 80/20 train-test split with scikit-learn.

I’ve also strengthened model reliability and deployment readiness by removing data leakage (removing 5 failure-mode indicator columns) and tuning performance with RandomizedSearchCV and 5-fold cross-validation (reported 0.67 validation macro F1-score). Alongside modeling, I’ve audited 60,000 records using a 4-phase SQL pipeline for data integrity, and I’ve supported analytics outputs via interactive Power BI dashboards and reproducible notebooks.

Beyond foundations, I’ve developed projects that emphasize integrity and communication—conducting parametric mediation analysis on 253,680 CDC BRFSS records using the MacKinnon Product-of-Coefficients framework and building an interactive Streamlit Cloud dashboard. I’m energized by causal reasoning, careful validation, and delivering results stakeholders can act on.