Project Overview
Accurately anticipating future demand is important for operational planning. For this project, I developed and compared time-series forecasting models to predict demand counts, identify seasonal behavior, and support capacity planning decisions.
Methodology & Modeling
Using JMP statistical software, I compared forecasting approaches for count data with recurring seasonal behavior:
- 1. ARIMA & SARIMA Models: Tested ARIMA models and selected a Seasonal ARIMA(1, 1, 1)(0, 1, 1)12 model for the validation comparison.
- 2. Seasonal Exponential Smoothing: Compared the SARIMA result against a 12-period Seasonal Exponential Smoothing model using the Zero to One option shown in the JMP output.
Results & Visualization
Both models captured meaningful variation in the Actual Count data. The Seasonal ARIMA model marginally outperformed the Seasonal Exponential Smoothing model, achieving an R-squared of 0.5909 compared with 0.5812.
Actual-by-predicted comparison from JMP validation output.
Operational Impact
By translating historical count data into forecast comparisons, this approach helps planners evaluate expected demand, prepare staffing or resource levels, and review model fit before using forecasts for operational decisions.