基于WOA-Stacking集成学习的注塑产品尺寸预测
Injection molding product size prediction based on WOA-Stacking ensemble learning method
陈忠杭 1王舟挺 2沈加明 2胡燕海 1倪德香2
作者信息
- 1. 宁波大学机械工程与力学学院,浙江宁波 315211
- 2. 宁波华美达机械制造有限公司,浙江宁波 315000
- 折叠
摘要
在现有的基于机器学习的注塑产品尺寸预测模型中,存在单一模型预测精度不高的问题,为了提高实时监测注塑产品尺寸变化的精度,提出了一种基于鲸鱼优化算法(WOA)优化Stacking集成学习的注塑产品尺寸预测方法.首先,整合注塑过程收集到的数据,使用3σ准则进行异常值筛选,再通过随机森林法和互信息法选取关键的特征,作为后续模型的输入特征;其次,在Stacking集成学习框架中,选择K近邻、随机森林和轻量级梯度提升机作为基学习器,选择弹性网络回归作为元学习器,使用WOA优化各个基学习器中的超参数,构建WOA-Stacking集成学习预测模型;最后,将所提的模型应用到注塑产品尺寸预测并与其他模型进行对比分析,以验证本方法的有效性.以第四届工业大数据创新竞赛数据为例,在包含3种集成模型和3种单一模型的对比实验中,选择产品的三维尺寸作为预测目标,实验结果表明WOA-Stacking集成学习模型具有更高的预测精度和拟合能力.
Abstract
In the existing machine learning-based injection molding product size prediction models,there is the problem that the prediction accuracy of a single model is not high.In order to improve the accuracy of real-time monitoring of injection molding product dimensional changes,an injection molding product size prediction method based on whale optimization algorithm(WOA)optimized Stacking ensemble learning was proposed.First,the data collected from the injection molding process were integrated,the outliers were screened using the 3σ criterion,and the key features were selected by random forest and mutual information method as the input features for the subsequent model.Second,in the Stacking ensemble learning framework,K-nearest neighbor,random forest and light gradient boosting machine were selected as the base learners,and elastic net regression was selected as the meta learner,and WOA was used to optimize the hyperparameters in each base learner to construct the WOA-Stacking ensemble learning prediction model.Finally,the proposed model was applied to the injection molding product size prediction,and then other models were compared and analyzed to verify the effectiveness of the present method.Taking the data of the 4th Industrial Big Data Innova-tion Competition as an example,in the comparison experiment containing three ensemble models and three single models,the three-dimensional size of the products were selected as the prediction target,and the experimental results showed that the WOA-Stacking ensemble learning model had higher prediction accuracy and fitting ability.
关键词
注塑/尺寸预测/鲸鱼优化算法/Stacking集成学习/特征选择Key words
injection molding/size prediction/whale optimization algorithm/Stacking ensemble learning/feature selection引用本文复制引用
出版年
2024
NETL
NSTL
万方数据