Forecasting Inventory Demand Under Volatile Sales Patterns Using the Prophet Algorithm

Inventory availability is a critical factor for companies to maintain operational continuity and customer satisfaction. However, many organizations still face challenges in forecasting demand, particularly when sales patterns are highly volatile and irregular. Although the Prophet forecasting algorithm has been widely used for time-series prediction, its behavior and robustness under unstable sales patterns remain insufficiently examined in practical inventory contexts. This study aims to evaluate the ability of the Prophet algorithm to forecast inventory demand using historical sales data characterized by fluctuating patterns. A quantitative time-series forecasting approach was applied using one year of secondary sales data obtained from PT XYZ. The data were cleaned to address missing values and aggregated into weekly time intervals to reduce noise. Five products with the highest transaction frequency were selected as case studies. Forecasting performance was evaluated using Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE). The results show that Prophet is capable of generating reasonably accurate forecasts even under volatile demand conditions. The evaluation results indicate RMSE values ranging from 5.41 to 52.78 and MAPE values ranging from 5% to 23.46% across the five analyzed products. These findings provide empirical evidence that the Prophet algorithm can maintain forecasting robustness despite irregular demand patterns. However, the absence of comparisons with alternative forecasting models limits the strength of conclusions regarding its relative performance. This study contributes by providing empirical insight into the application of Prophet for inventory forecasting under volatile sales conditions and offers practical implications for improving inventory planning in data-driven decision-making environments.