Clustering and Forecasting Beef Prices: WPGMA and Time Series Regression Integration

Introduction. Beef is the second most consumed animal protein in Indonesia after chicken. Its demand increases during Eid al-Fitr and Eid al Adha, driving seasonal price spikes. These fluctuations challenge the government in maintaining price stability, with regional differences further complicating policy decisions. Objectives. This study aims to group Indonesian provinces by monthly beef price trends using time series clustering and to perform group-level forecasting with Time Series Regression that incorporates the effects of Eid al-Fitr and Eid al Adha. This approach simplifies modeling complexity and supports efficient policy-making for beef price stabilization. Materials and Methods. Principal Component Analysis (PCA) was used to optimize the dataset before clustering using the Weighted Pair Group Method with Arithmetic Mean (WPGMA) algorithm, combined with the Autocorrelation-Function (ACF) Distance similarity measure. Time Series Regression (TSR) models with dummy variables for Eid al-Fitr and Eid alAdha effects were then applied to each cluster for forecasting. Results. The analysis identified five optimal clusters, with a silhouette coefficient value of 0.61. Forecasting within each cluster showed in sample MAPE values ranging from 1.05% to 5.15%. All clusters exhibited an increasing trend in beef prices from September 2023 to August 2024, highlighting the impact of calendar events and regional characteristics on price dynamics. Conclusions. The clustering-based forecasting approach effectively simplifies price analysis across regions, supporting targeted policy interventions and improving the accuracy of beef price predictions in Indonesia.