Demand Forecasting on a Mobile-Based Inventory Application: A Case Study of Jaya Sentosa Store

Limited customization in existing inventory manage-ment systems hinders optimal stock management for businesses with unique needs. This study addresses this gap by developing a mobile-based inventory application specifically designed for Jaya Sentosa Store. The application utilizes demand forecasting to optimize restocking decisions. While various demand forecasting models exist (e.g., ARIMA, SARIMAX), selecting the most suitable model for a specific business can be challenging. Our research implements and compares the accuracy and runtime of several models (ARIMA, SARIMAX, Theta, Prophet, LSTM) to determine the optimal choice for Jaya Sentosa Store. We apply Scrum in the development process. Back-end and front-end development use JavaScript (JS) and Kotlin programming languages, respectively. We conduct several experiments to de-termine the most suitable forecasting model between ARIMA, SARIMAX, Theta model, Prophet, and LSTM. The test uses MAE (Mean Absolute Error), RMSE (Root Mean Squared Error), and runtime as parameters. When it comes to accuracy and consistency, the Theta model excels in the test. Besides its runtime score, Theta model has the best average MAE and RMSE scores. Theta model has an average runtime, MAE, and RMSE of 0.053, 3.951, and 3.226, respectively. We conduct black box testing on the inventory application with Jaya Sentosa Store where all the cases are valid.