基于LightGBM模型的离散制造业产品物料需求智能预测

Intelligent forecast of material demand for discrete manufacturing products based on LightGBM model

离散制造业产品的物料需求受多种因素影响,传统物料需求预测算法对数据要求高,企业需要进行大量运算,且预测提前期短、精准度低,不能及时满足企业的生产计划。为提高制造业物料需求预测精度,本研究采用美的集团离散型物料需求数据进行建模分析。首先进行数据预处理及特征工程,利用统计学中的统计量构建出滑动和滞后特征,然后构建并拟合LightGBM模型对物料需求量进行预测,并与传统时间序列SARIMA模型进行对比,引入平均绝对误差MAE评估模型的预测精准度,针对模型时间复杂度和预测精准度进行对比分析。结果表明,以月为时间粒度的情况下,LightGBM机器学习模型对离散制造业物料需求预测的效率和准确率更高,更有利于提高离散制造企业的生产效率。

The material demand of discrete manufacturing products is influenced by various factors.Traditional material demand prediction algorithms require high data requirements,and enterprises need to perform many calculations.Moreover,the prediction lead time is short,and the accuracy is low,which cannot meet the production plan of enterprises in a timely manner.To improve the accuracy of material demand prediction in the manufacturing industry,this study uses discrete material demand data from Midea Group for modeling and analysis.First,data preprocessing and feature engineering are carried out,and the statistics in statistics are used to build the sliding and lagging characteristics.Then,the LightGBM model is built and fitted to forecast the material demand and compared with the traditional time series SARIMA model.The prediction accuracy of the average absolute error MAE evaluation model is introduced,and the model time complexity and prediction accuracy are compared and analyzed.The results indicate that the LightGBM machine learning model has higher efficiency and accuracy in predicting material demand in discrete manufacturing industries with monthly time granularity,which is more conducive to improving the production efficiency of discrete manufacturing enterprises.