<![CDATA[The dynamic development of the business world demands that Micro, Small, and Medium Enterprises (MSMEs) have an accurate and easy-to-use sales forecasting system, considering that many MSMEs still rely on intuitive methods that cause supply chain inefficiencies. This study aims to design a prototype of a multi-method-based sales forecasting system that is adaptive to data characteristics and can be easily operated by MSMEs with various levels of digital literacy. The methodology used includes collecting and analyzing user needs, data pre-processing (including outlier detection and ABC classification), and time series modeling with five methods (ARIMA, Holt-Winters, Prophet, Theta, LightGBM). Model accuracy was evaluated using Mean Absolute Percentage Error (MAPE) and Root Mean Squared Error (RMSE) on data divided 70:30 for the training set and test set, while usability was assessed using the System Usability Scale (SUS) with ten non-technical respondents. The test results showed that the system was able to automatically adjust the selection of the best method based on historical product data patterns. The system also achieved an average SUS score of 90, indicating an "excellent" level of usability. Therefore, this prototype has the potential to be a practical sales prediction tool for MSMEs to support data-driven decision-making.]]>
Copyrights © 2025
