Fares Fahim Bechir Gbetnkom

I've built end-to-end machine learning projects for inventory analytics, returns classification, and energy-price forecasting, using real data and clearly disclosed business simulations.

My Project SIGMA pipeline segmented 4,730 SKUs into four risk tiers using PCA and Gaussian Mixture Models, projecting more than $700K in annual value through cluster-specific strategies. For ReturnX, I developed a TF-IDF, XGBoost, and SMOTE classifier that achieved a 0.93 Macro F1-score and routed low-confidence cases to human review. For PJM Energy Price Forecasting, I trained a GradientBoosting model on one year of hourly data, achieving a 16.3 CV RMSE and outperforming six baselines, with a simulated $202K profit from a $428K theoretical maximum. At Smartrack Africa, I validated vehicle-tracking data and built xSAVE, a digital archiving system that improved record management and data integrity.