<![CDATA[Inaccurate forecasting of stock requirements and market demand is a major challenge faced by PT Bintang Jaya, which may lead to excess inventory (overstock) or stock shortages (stockout). This issue occurs because the company’s stock planning process still relies on manual approaches and historical experience without optimal utilization of data analytics. Therefore, this study aims to apply machine learning–based prediction techniques to estimate stock needs and market demand more accurately. The methods used in this research include the Random Forest algorithm as the baseline model, and Random Forest combined with Principal Component Analysis (PCA) as a hybrid model to evaluate the impact of dimensionality reduction on prediction performance. The dataset consists of historical sales transaction records from PT Bintang Jaya during the 2022–2024 period, which were processed through data preprocessing, monthly aggregation, and time series feature engineering. The results show that the Random Forest model provides more stable demand predictions and is closer to the actual values compared to the hybrid RF+PCA model. The application of PCA did not improve prediction performance due to the characteristics of the dataset, which is relatively low-dimensional and non-linear. Overall, the baseline Random Forest model demonstrates good and stable performance, indicated by consistent MAE and RMSE values and a coefficient of determination (R²) of approximately 0.69, meaning that the model explains around 69% of the demand variation based on the historical features.]]>
Copyrights © 2026
