Real-Time Production Forecasting Using GSS-Optimized Holt DES: An Empirical Comparison with ARIMA

Kurnia Plastic Factory depended on manual documentation of transactions and production scheduling, leading to postponed reporting, erroneous completion projections, and inadequate decision-making. This research created an online point-of-sale system featuring automated predictions for production duration and assessed its forecasting precision and computational efficacy in comparison to a statistical standard. The system was constructed utilizing the Extreme Programming style, the Laravel framework, PHP, and MySQL. Holt Double Exponential Smoothing technique was refined through the application of the golden section search algorithm. Functional testing verified that all 19 features functioned as intended, whereas user acceptance testing attained a success percentage of 94.74%. The assessment of forecasting efficacy was conducted by rolling one-step-ahead projections utilizing 255 observations of production duration documented from 2023 to 2025. The autoregressive integrated moving average model with parameters of one, one, and one surpassed the optimum smoothing technique in all accuracy measures, recording a mean squared error of 1.69, a mean absolute error of 1.05, and a mean absolute percentage error of 34.99%. The refined smoothing technique necessitated an average processing duration of 2.36 milliseconds, in contrast to 35.77 milliseconds for the fixed-order statistical model. The prediction endpoint yielded a response time of about 154.20 milliseconds. The smoothing technique was preserved due to its automatic parameter adjustments for each request, resulting in reduced processing expenses, even though it exhibited diminished forecasting precision. These results revealed a compromise between predictive precision and computational efficacy in production scheduling.