Publication: Daily retail demand forecasting of bakery products using machine/ deep learning-based algorithm
Loading...
Date
2024-07
Authors
Easter, Wong Poh Yee
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This thesis investigates the application of SARIMA and XGBoost models for forecasting daily demand for bakery products across three stores: Sabasun, Xiri, and
Pasar Raya Seni Intan. By evaluating and comparing the performance of these models on both smoothed and unsmoothed datasets, the study aims to enhance forecast accuracy. Key evaluation metrics such as MAPE, MAE, MSE, and RMSE were used to assess the models. The analysis revealed that XGBoost generally outperformed SARIMA in terms of accuracy and robustness, effectively handling both smoothed and unsmoothed datasets. In contrast, SARIMA struggled with the volatility present in the unsmoothed datasets, leading to larger prediction errors. High MAPE values in both models suggest significant data variability and the presence of outliers, impacting forecast accuracy. Despite these challenges, the thesis achieved its objectives: reducing MAE to below 5.0 and RMSE to below 31.0 using XGBoost, which demonstrated MAE ranging from 0.0041 to 2.0978 and RMSE from 0.1112 to 3.5450. These findings underscore the effectiveness of the selected models and preprocessing methods, recommending XGBoost for its superior performance and adaptability. However, the high data variability indicates the need for further refinement and exploration of additional or hybrid models to enhance forecasting accuracy.