Penerapan Data Mining dalam Peramalan Permintaan untuk Pemesanan Stock Mobil di PT. XYZ

Abstract

The advancement of information technology has driven the automotive industry, including PT. XYZ, to utilize data in making strategic business decisions. One of the main challenges faced is the high volume of vehicle indent orders due to inaccurate ordering from dealers to the main dealer. This study aims to forecast the demand for Honda vehicles (Brio, WRV, BRV, HRV) to reduce indent occurrences and improve stock management efficiency. The methods used include K-Means Clustering for data segmentation based on specific characteristics, followed by demand forecasting using Random Forest and K-Nearest Neighbor algorithms. Model performance was evaluated using Mean Absolute Percentage Error (MAPE) and R-Squared (R²). The results show that the Random Forest model provided higher forecasting accuracy than K-Nearest Neighbor, with an average MAPE of 26.31% and R² = 0.83, lower than K-Nearest Neighbor's average MAPE of 55.18% and R² = 0.29 across most clusters. In conclusion, the data mining approach applied in this study effectively identifies demand patterns and provides strategic recommendations for vehicle ordering to prevent imbalances between demand and inventory.