Production Planning Forecasting System Based on M5P Algorithms and Master Data in Manufacturing Processes

Abstract

With the increasing adoption of smart factories in manufacturing sites, a large amount of raw data is being generated from manufacturers’ sensors and Internet of Things devices. In the manufacturing environment, the collection of reliable data has become an important issue. When utilizing the collected data or establishing production plans based on user-defined data, the actual performance may differ from the established plan. This is particularly so when there are modifications in the physical production line, such as manual processes, newly developed processes, or the addition of new equipment. Hence, the reliability of the current data cannot be ensured. The complex characteristics of manufacturers hinder the prediction of future data based on existing data. To minimize this reliability problem, the M5P algorithm, is used to predict dynamic data using baseline information that can be predicted. It combines linear regression and decision-tree-supervised machine learning algorithms. The algorithm recommends the means to reflect the predicted data in the production plan and provides results that can be compared with the existing baseline information. By comparing the existing production plan with the planning results based on the changed master data, it provides data results that help production management determine the impact of work time and quantity and confirm production plans. This means that forecasting data directly affects production capacity and resources, as well as production times and schedules, to help ensure efficient production planning.