Forecasting Demand for Donuts Using a Machine Learning Approach

UKM Donat XYZ is a company focused on producing donuts with a daily capacity of 4,000 to 5,000 units, offering more than 20 flavor variants. However, the company experiences monthly sales fluctuations and often encounters issues with unsold donut stock due to difficulties in accurately predicting production volumes manually. The research was conducted to forecast donut demand in 2024 using a machine-learning approach to address this challenge. The study leveraged historical daily sales data from 2022 and 2023, totaling 730 data points. Two methods were employed: the Naïve-Bayes algorithm and Prophet. Evaluation of the results indicated that the Naïve- Bayes algorithm achieved better accuracy than Prophet, with a Mean Square Error (MSE) of 428,328.92 for Naïve-Bayes. At the same time, Prophet had an MSE of 27,178,280. This research provides valuable insights for the company to efficiently manage its production and optimize donut stock. By utilizing predictive models like Naïve-Bayes, the company can better forecast market demand and avoid situations of potential stock shortages or excesses that could lead to losses. It also aids in more accurate production planning and efficient resource management. This research is anticipated to serve as a foundation for the company to enhance its strategies in meeting consumer demand more timely and effectively. Consequently, UKM Donat XYZ can improve its operational performance and reduce potential losses from wasted products due to errors in demand estimation.