Prediction of Raw Material Usage for Hijab Production Using k-Nearest Neighbors and Random Forest Methods

Accurate demand forecasting will optimize the supply chain management in the hijab industry, since demand forecasting is highly required for maintaining efficient and effective production and inventory processes. This study investigates the k-Nearest Neighbors ($k-\mathrm{NN}$) and Random Forest (RF) for the purpose of predicting demands regarding raw materials in the hijab production process. Key performance metrics considered for evaluating both models include the RMSE, MAE, MAPE, and R2. These findings indicate that k-NN outperforms RF in terms of superior prediction accuracy with $K=2$. In this respect, the root mean square error from k-NN was 1.768 against the RF’s RMSE of 1.801, proving that it is always more reliable in capturing local variations in data. On top of this, even incomplete data or those occasions when demand shifts all over the place were handled well by k-NN with more consistent and thus more accurate results. These results offer valuable insights for hijab manufacturers, allowing them to optimize their supply chain management by optimizing material shortages and overstock inventory. This research further highlights the importance of choosing the right predictive model according to the context and data characteristics to improve operational efficiency and enhance customer satisfaction.