Using Time Series Models in Product Based Order Forecasting

Abstract

Production systems play a vital role in maximizing consumer satisfaction by efficiently transforming inputs such as labor, raw materials, and capital into products or services aligned with consumer demands. An order-based production takes place in poultry meat and meat products production facilities, which face various difficulties in meeting changing customer demands and managing the supply of raw materials. To optimize production and increase customer loyalty, these facilities use strategic scheduling, considering their daily production capacity and fluctuating customer orders. In this study, estimating which customer and product type the future order quantities will come from for the relevant facilities, increasing customer satisfaction by facilitating order processes and minimizing storage costs are discussed. With this study, the number of orders was estimated, and it was aimed to meet the orders in the most accurate way. In the estimations, the order data of a poultry meat and meat products production facility between 2013 and 2021 were used. Since the order figures will change every year in cases such as the customer working with the facility, growing, or shrinking, better results have been tried to be obtained with the arrangements made on the data set used and three different data sets have been obtained. Estimation processes were performed for these three data sets using LSTM and Prophet algorithms. While the RMSE value was 7.07 in the LSTM model in experimental studies, this value was obtained as 10.96 for Prophet. In the results obtained, it was observed that the arrangements made on the data set positively affected the accuracy of the estimations and the LSTM algorithm produced better results than the Prophet algorithm.