DATA 624: Predictive Analytics

Summer 2026 — Group 3 Portfolio

This portfolio collects the group work submitted for DATA 624: Predictive Analytics at CUNY SPS, Summer 2026. It covers time series forecasting, predictive modeling, and model evaluation across two homework assignments and two group projects.

Group 3

Masoud Mahdisoltani, Andre Thomson, Desiree Thomas, Kevin Martin, Khandker M Qaiduzzaman

---

Homework 1 applies foundational time series concepts from Hyndman and Athanasopoulos (FPP3) alongside preprocessing and EDA exercises from Kuhn and Johnson (KJ), covering decomposition, ETS, ARIMA, and predictor transformations.

Homework 2 extends into supervised learning with KJ exercises on regularization, tree-based models, gradient boosting, and Cubist, plus a market basket analysis using association rules.

Project 1 delivers production forecasts for two real clients: quarterly shipment demand for Lone Star Appliances and hierarchical patient demand across five Texas HealthLink clinics. Includes a Python vs. R forecasting comparison and an interview-ready time series literacy section.

Project 2 builds and evaluates a suite of regression models to predict beverage pH from sensor and process data, with Cubist selected as the champion model. Includes a Python vs. R modeling comparison using scikit-learn.