Conducted a customer churn analysis simulation for XYZ Analytics (BCG Forage). Leveraged Python (Pandas, NumPy) for Exploratory Data Analysis (EDA) and built an 85% accurate Random Forest model. Delivered actionable recommendations to non-technical stakeholders, showcasing clear communication and strategic thinking to reduce customer attrition. Tools: Python, Pandas, NumPy, Scikit-learn (Random Forest), EDA.
As part of my Business Data Mining coursework at Trinity College Dublin, I developed a Random Forest model to predict remote work productivity. This project involved data collection (3,000+ survey responses), preparation, feature engineering, EDA, and model interpretation, achieving 83.39% accuracy and providing actionable insights. Tools: Python, Scikit-learn, Pandas, NumPy, EDA.
Conducted in-depth analysis of customer behavior and sales trends for the Google Merchandise Store using SQL queries in BigQuery and Tableau for visualization. Developed data-driven recommendations to optimize marketing strategies, reduce checkout abandonment, and enhance customer retention. Tools: SQL (BigQuery), Tableau, Data Analysis.
Analyzed the impact of weather, pests, and geographical factors on pineapple crop health for Fyffes. This involved data integration, trend analysis, and predictive modeling to develop insights on disease prevention, root health, and proactive intervention strategies to mitigate crop losses. Tools: Data Analysis, Predictive Modeling, Python, Power BI, Excel.
Proposed a Big Data and AI strategy for Paddy Power to optimize customer retention, fraud detection, and responsible gambling. This included suggesting machine learning models (e.g., for churn prediction, risk management) and personalized promotions, outlining a data-driven framework for enhanced engagement and compliance. Tools/Concepts: Big Data Strategy, AI, Machine Learning, Churn Prediction, Risk Management.
Led an analytical project investigating ESG rating discrepancies in Chinese companies. This involved integrating Bloomberg ESG scores, independent ESG ratings, and news sentiment data into a relational SQLite database.
I then developed predictive models (XGBoost, Random Forest, Linear Regression) to quantify ESG decoupling and applied time-series forecasting (ARIMA) to assess firm-level trends. Key findings revealed a significant average decoupling rate of 60.72% and highlighted a weak alignment between news sentiment and reported ESG scores.
Tools: Python (Pandas, Scikit-learn, XGBoost, Statsmodels), SQLite, Power BI, Tableau.