I love solving problems using data engineering tools
dealing with large dataset
Where business and healthcare problems are solved with the power of data science. Dive into my portfolio and explore real-world machine learning solutions. Come over to my GitHub here.
I am Clement, a passionate data professional with a knack for transforming complex data into impactful insights.
From data pipelines to machine learning, I bring structure, clarity, and intelligence to data-driven decisions.
My toolkit spans the full data science spectrum — from querying databases and building dashboards to training robust ML models. Below are just a few things I bring to the table.
Check out some of my highlighted projects covering data engineering, machine learning, analytics, and visualization. Each project is linked directly to the full GitHub repo so you can explore the code and insights yourself.
Leverage K-Means and Hierarchical Clustering to discover unique customer segments in an online retail dataset. This unsupervised learning approach provides data-driven insights for personalized marketing, improved customer retention, and smarter credit strategies.
This supervised learning project utilizes Decision Trees and Random Forest classifiers to predict whether clients will subscribe to a term deposit based on bank marketing campaign data. The solution includes complete preprocessing, feature engineering, hyperparameter tuning, and model evaluation. 🔄 The project was also replicated and validated using Azure ML Studio, enhancing model interpretability and reproducibility within a cloud-based environment.
Analyze thousands of recipe reviews using VADER and Naive Bayes to detect customer sentiment and link it with star ratings. Uncover what customers love or dislike, and visualize it with word clouds and sentiment trend charts.
This project enhances traditional Marketing Mix Modeling by integrating machine learning techniques like XGBoost and Random Forest to predict revenue and optimize marketing spend. SHAP values support multi-touch attribution, while SARIMA handles long-term forecasting. Budget reallocation and optimization were performed using simulations and constrained programming. Built with Python and Azure ML Studio, the framework offers interpretability, accuracy, and actionable insights, achieving a 96% R² in predictive performance.
Advanced regression techniques including multi-linear regression, Random Forest, and XGBoost were used to model concrete strength based on its ingredients. Log-transformations and diagnostics ensured assumptions were met, with XGBoost yielding R² of 0.997.
Modeled stock prices using ARIMA, Holt-Winters, and SARIMA techniques. Forecast accuracy was validated using ACF, Ljung-Box, RMSE, and MAPE. The SARIMA model emerged as most effective for capturing both trends and seasonality.
Designed and normalized a relational database into 3NF for a hospital. Built T-SQL procedures, constraints, views, and triggers for handling patients, appointments, diagnoses, feedback, and backups. Emphasized data integrity, concurrency, and access control security.
Designed an interactive dashboard analyzing urban and rural population growth globally from 1960 to 2050. Utilized DAX measures, slicers, and drill-down hierarchies for real-time data exploration. Ideal for policy insights and regional planning.
Used RDDs, Spark SQL, and DataFrames to process clinical trial datasets across multiple years. Executed complex data cleansing, sponsor filtering, and aggregation pipelines with visualizations in Matplotlib and Power BI.