KR-20260065418-A - Optimal work order solution through AI demand forecasting

Abstract

The present invention relates to a delivery date prediction algorithm for standard steel products. By utilizing various inventory and production management information recorded and stored in a database to predict possible delivery dates for standard steel products received from customers, the invention automates the delivery date calculation method that previously relied on existing worker experience, thereby providing a more rapid and accurate prediction. This enables the quick and easy confirmation of orders, making it a highly useful invention that facilitates rapid order processing and increased productivity.

Inventors

• 박상준
• 한상렬

Assignees

• 데이터플라이 주식회사

Dates

Publication Date

20260508

Application Date

20241101

Claims (2)

1. How to redefine overpayment, inventory, and production data affecting the delivery of ordered products
2. An AI delivery date prediction algorithm that predicts delivery dates by inputting the data redefined in Clause 1 into a delivery date prediction algorithm based on Gradient Boosting Machine (GBM).

Description

Optimal work order solution through AI demand forecasting Optimal work order solution through AI demand forecasting The present invention relates to a delivery date prediction algorithm that predicts the delivery date for orders of standard steel products using a Gradient Boosting Machine (GBM) algorithm utilizing inventory and production data, and enables quick and easy order acceptance after collecting and analyzing receipt data, production, and inventory data generated upon order acceptance. By predicting delivery dates in this way, an optimal production plan can be created, and when workers utilize this plan to carry out their work, the efficiency of inventory, sales, and production can be increased. The surge in demand resulting from the revitalization of the standard steel material market has increased the need to shorten and manage production and delivery times. However, since most processes—from the procurement of steel raw materials to process and inventory management and the handling of customer requests—are currently handled manually, continuous errors are occurring. Consequently, it is difficult to estimate delivery dates to customers, with delays averaging more than 15 days per month being the norm. Therefore, there is an increased need for accurate and rapid decision-making through precise delivery schedules and production instructions in response to customer orders. For example, when a customer orders a standard steel product, the inventory, outstanding quantity, and production status of the product corresponding to the material and dimensions are checked one by one, and then the delivery date of the requested standard steel product is predicted and notified to the customer. Since the requested order cannot be promptly confirmed due to manual processing, it lowers the customer's purchasing motivation, resulting in a problem where performance cannot be efficiently increased. Accordingly, the present invention enables the calculation of delivery dates—which was previously performed manually using expert know-how—to be carried out more quickly and accurately through the calculation of delivery dates using an algorithm (delivery date prediction). Figure 1 shows inventory and production data redefined to be suitable for delivery date forecasting. Figure 2 illustrates the process of an algorithm that calculates the expected delivery date through a delivery date prediction algorithm for standard steel material products. Hereinafter, the structure of the present invention and the resulting operation and effects will be described collectively with reference to the attached drawings. To calculate the delivery date, inventory and production data are collected and processed. Subsequently, in the actual delivery date forecast, the data is redefined by reflecting the expert's delivery date calculation conditions as shown in Fig. 1 and loaded into a database. At this time, the redefined data newly includes information other than the order quantity, such as the quantity overdue as of the previous day, inventory considering processable thickness/width, and the order day (to account for public holidays). Then, the delivery date is predicted using the Gradient Boosting Machine (GBM) algorithm. To achieve the optimal delivery date prediction with the smallest error, the optimal algorithm was selected by adjusting hyperparameter values such as depth, tree, and cascade stage. The optimal estimated delivery date can be calculated using the following AI algorithm, and the delivery date prediction algorithm process for standard steel material products can be verified by referring to Fig. 2.