Abstract:The time series data of material demand of manufacturing companies are often non-stationary. Paper uses empirical mode decomposition (EMD) to convert non-stationary time series into a series of intrinsic mode function (IMF) and a residual term (RES), and then digged out more information combined with least squares support vector machine regression (LSSVR) to forecast the model. Finally, the empirical results show that the EMD-LSSVR combination forecast can effectively predict non-stationary material demand time series, and the prediction accuracy is high. It has a certain degree of promotion and practical value. Introduction In recent years, how to accurately predict changes in demand information to avoid the generation of bullwhip effect has become a hot topic in academia. At the same time, customers demand for timely response to the urgent attention of the market environment. Companies improve their ability to forecast material requirements that is a powerful guarantee for them to gain competitive advantages. In a customer-oriented market environment, on the one hand companies need to respond quickly to customer needs; on the other hand, companies also want to be able to limit output and minimize procurement and inventory costs [1]. However, due to the existence of bullwhip effect in the supply chain system, changes in demand will increase as the magnitude of the supply chain level increases, resulting in problems such as the lack of timely response of the demand or the surge of inventory. Therefore, for each enterprise in the supply chain, it is of great significance to accurately predict the actual demand of materials. With regard to research on demand forecasting methods, domestic and foreign scholars have obtained certain research results and classified the forecasting methods into qualitative analysis methods and quantitative analysis methods. Qualitative analysis mainly includes research methods, judgment analysis methods, Delphi expert survey methods and customer comprehensive judgment method; Quantitative analysis method is mainly based on historical data using modern mathematical methods to analyze and process historical data by establishing corresponding forecasting models to analyze, which mainly including arithmetic average method, moving weight method, exponential smoothing method, regression analysis method, grey prediction method and so on. For the manufacturing industry, the material demand is affected by multi-dimensional factors. For the general forecasting model, it is necessary to know in advance the model of the influencing factor. This model mainly depends on subjective analysis and feature discrimination, so the model is not ideal [2]. In 1998[3], Dr. Huang E proposed a data-based adaptive prediction method called Empirical Mode Decomposition (EMD). This method can reveal the hidden patterns and trends of time series, simplify the prediction task into several simple prediction subtasks, and can be applied to various complex prediction problems. The advantage of the EMD method is that it is well-suited for decomposing nonlinear and non-stationary time series, and performs better than wavelet and Fourier decomposition in describing local time instantaneous frequencies. International Conference on Education Reform and Management Science (ERMS 2018) Copyright © 2018, the Authors. Published by Atlantis Press. This is an open access article under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/). Advances in Social Science, Education and Humanities Research, volume 177

Prediction and analysis for dynamic trend of customer requirements weight on product

Weight Prediction for Technical Characteristics in QFD Based on LS-SVM

Dynamic Weighted Least Squares Support Vector Machines

Study on Dynamic Weighing System Base on Intelligent Algorithm

Support Vector Machines-Based Method for Machining Error Prediction Modeling

Analysis and prediction for dynamic requirements in house of quality based on grey theory

Dynamic Job Shop Scheduling Based on Order Remaining Completion Time Prediction

Research on Long Short-Term Decision-Making System for Excavator Market Demand Forecasting Based on Improved Support Vector Machine

Product Quality Improvement Method In Manufacturing Process Based On Kernel Optimisation Algorithm

Assistant Fashion Design Model Based on Markov Prediction

A Hybrid Support Vector Machines and Logistic Regression Approach for Forecasting Intermittent Demand of Spare Parts

Multivariate SVR Demand Forecasting for Beauty Products Based on Online Reviews

Prediction of Standard Time of the Sewing Process using a Support Vector Machine with Particle Swarm Optimization

Dynamic Model Selection Based on Demand Pattern Classification in Retail Sales Forecasting

Material Demand Combination Forecasting Model Based on EMD-PSO-LSSVR

A Layered KNN-SVM Approach to Predict Missing Values of Functional Requirements in Product Customization

Adaptability of SVR Time Series Analysis Used in Forecasting of Logistics Demand

Dynamic elicitation and forecasting innovation requirement of smart product-service system via user-manufacturer value co-creation perspective using multi-source data

Predicting total sales volume interval of an experiential product with short life cycle before production: similarity comparison in attribute relationship patterns

Research on a Dynamic Algorithm for Cow Weighing Based on an SVM and Empirical Wavelet Transform

Developing a Generalized Regression Forecasting Network for the Prediction of Human Body Dimensions