Customer Churn Prediction
BAN 614 — Machine Learning | University of Dayton | 2022
Predicted customer churn for a telecom company using 6,999 CRM records and 20 predictor variables. Built and compared 7 classification models.
-
Best model: Tuned Random Forest (~82% accuracy)
-
- Key finding: Fiber optic customers churn at 1.8x the rate of non-fiber; 35% of predicted churners have less than 6 months tenure
-
- Business output: $530K annual revenue at risk identified; proposed 25% discount intervention with projected +$85K/year upside
-
- Methods: Logistic regression (10-fold CV), LDA, Naive Bayes, decision trees, bagging, random forest; ROC/AUC evaluation
NCAA Basketball Data Warehouse
MIS 664A — Data Management | University of Dayton | 2023
Designed and built a dimensional data warehouse for the 2022 NCAA men's basketball season from scratch using a live API data source.
-
Built full schema: DIM_TEAM, DIM_GAME, DIM_PLAYER, DIM_STADIUM, DIM_DATE, FACT_TEAMGAME, FACT_TEAMSEASON
-
Wrote complete DDL, DML, ETL documentation, and analytical query library, including complex 6-table joins
Inventory Optimization — Copeland
University of Dayton MBAN Capstone | Best Capstone Award | 2024
Non-linear optimization model built for Copeland's distribution center. Goal: improve RDSL (Request Date Service Level) while reducing FY2024 inventory value.
-
Team leader and primary client contact (Copeland Materials and Distribution team)
-
Built optimization model minimizing total inventory value subject to fill rate and safety stock constraints
-
Delivered root cause analysis showing 80% of late shipments occurred when inventory was physically on hand — a fulfillment process problem, not purely an inventory problem
-
-Delivered interactive Tableau visualization tool, sensitivity analysis, and item-level recommendations to Copeland leadership
-
Won Best Capstone Project across the entire MBAN program in 2024
**Tools:** Python (PuLP, OR-Tools), R, Tableau, Excel