Design Of Artificial Neural Network (ANN) Based Analytics Model for Safety Stock and Reorder Point Prediction at Galeri 1 ITERA

Guido Immanuel Simanungkalit (Institut Teknologi Sumatera)
Hersa Dwi Yanuarso (Institut Teknologi Sumatera)
Frieska Ariesta Syafnijal (Institut Teknologi Sumatera)
Eka Rachmadi Endarta Putra (Institut Teknologi Sumatera)
Sherin Ramadhania (Institut Teknologi Sumatera)

Jurnal Manajemen dan Teknologi Rekayasa
Vol 5, No 2 (2026)

Inventory management in retail environments such as Galeri 1 ITERA faces significant challenges due to fluctuating and irregular demand patterns, which often lead to stockout and overstock conditions. This study proposes an integrated inventory analytics system that combines ABC classification, demand forecasting using Artificial Neural Network (ANN), safety stock (SS) and reorder point (ROP) calculation, as well as interactive dashboard visualization to support data-driven inventory decision-making. The research methodology consists of: (1) ABC analysis to determine inventory control priorities for 56 products; (2) demand forecasting using ANN models trained on historical sales data; (3) calculation of safety stock and reorder point based on forecasting results and lead time variability; and (4) development of an analytics dashboard using Microsoft Power BI. The proposed ANN model utilizes a multilayer feedforward architecture capable of capturing nonlinear demand patterns. Forecasting performance was evaluated using WAPE, RMSE, and MAE metrics. The results indicate that 24 category-A items (42.9%) contribute to 79.45% of the total inventory consumption value, highlighting the need for tighter inventory control. The ANN forecasting model achieved an average accuracy of 88.72% with an average WAPE of 11.28%, while most products showed stable MAE and RMSE values below 2.5 units. Forecasting outputs were subsequently integrated into safety stock and reorder point calculations, producing dynamic inventory control parameters ranging from 3–209 units for safety stock and 3–296 units for reorder points. The main scientific contribution of this study lies in the integration of ABC classification, ANN-based forecasting, and dashboard-based predictive analytics into a unified inventory management framework for retail operations. The resulting dashboard enables interactive monitoring of inventory conditions, stock alerts, demand predictions, safety stock, and reorder point values, thereby improving the effectiveness and responsiveness of inventory control decisions.