PERAMALAN DATA PENJUALAN UNTUK MENGETAHUI TINGKAT PENJUALAN MAKANAN HEWAN PELIHARAAN MENGGUNAKAN METODE DECISION TREE REGRESI

Abstract

The pet food industry has grown rapidly alongside the rising number of pet owners, intensifying competition among stores and making sales-data-driven inventory management increasingly essential, particularly at Pet Shop Marendal, which had only recorded transactions without using them for forecasting. This study aims to develop a forecasting model for pet food sales levels and to identify the factors that most influence them, in order to support stock planning and business decision-making. The method used is Decision Tree Regression with the CART algorithm, applied to 1,500 transaction records from January–June 2025, split with an 80:20 ratio and validated through 5-fold cross-validation. The results show that the model achieves very high accuracy, with an R² of 0.9963 and an RMSE of Rp49,168, while the variables Quantity Sold and Unit Price together account for 98.99% of feature importance. It is concluded that Decision Tree Regression is effective and reliable for short-term sales forecasting, with practical implications providing a basis for stock management strategies and product-mix optimization for pet food business owners.