Forecasting Teaching Materials Using the Autoregressive Integrated Moving Average (ARIMA) Method Case Study: Universitas Terbuka

Universitas Terbuka (UT) is an open and distance learning university in Indonesia, where teaching materials serve as the primary learning resources. Currently, UT relies on manual calculations using Microsoft Excel to estimate the required number of printed materials. This approach often results in inaccurate forecasts, leading to either surplus or shortage of teaching materials. Such discrepancies negatively impact operational efficiency and cost-effectiveness. Repeated duplication increases unit costs, while surplus materials risk expiration and waste due to content updates or limited shelf-life. This study applies a machine learning approach, specifically the ARIMA method, within the CRISP-DM framework to forecast teaching material needs more accurately. Results show that the ARIMA (2,1,0) model, with parameters p=2, d=1, q=0, and AIC=34.512 and BIC=30.591, provides the best performance. Teaching material EKMA4434, used in 11 study programs, recorded the lowest RMSE of 5,909.881 units, indicating an average prediction error of that magnitude against actual usage.