Research on Material Demand Forecasting Algorithm Based on Multi-Dimensional Feature Fusion

Introduction

Demand forecasting is an important basis for enterprises to formulate strategic planning, production arrangement, sales plan, and logistics management plan (Moscoso-López et al., 2016). For a business to efficiently manage its production, inventories, supply chain, finances, and market position, demand forecasting is a crucial tool. Businesses can make decisions that improve operational performance and boost profitability by precisely estimating demand. Manufacturers of standard products are expected to produce a certain amount of products ready for market or, at least, to keep a sufficient amount of raw materials and spare parts in order to minimize delivery time. Accurate material demand prediction can not only reduce inventory, but also ensure the normal production of enterprises, and effectively prevent production accidents, procurement accidents, and other material shortages caused by suppliers (Waller & Fawcett, 2013).

In recent years, many scholars have conducted extensive research on demand forecasting. FeiFei Ming et al. (2020) predicted the material distribution time in the production logistics system of the assembly workshop by establishing backpropagation (BP) neural network. Dong Jiang et al. (2019) used BP neural network to forecast engine material requirements and achieved good results. Zhou Yangfan et al. () studied the application of deep learning in logistics inventory prediction; the error reverse transmission function can be used in the prediction model.