基于LSTM深度学习模型的羽绒材料价格预测研究  

Study on the Price Prediction of Down Material based on Depth Learning Model of LSTM

作　　者：韦玉辉 唐欣[1] 许仲童 丁雪梅 吴开明[4] WEI Yu-hui;TANG Xin;XU Zhong-tong;DING Xue-mei;WU Kai-ming(College of Textile and Clothing,Anhui Polytechnic University,Wuhu Anhui 241000,China;College of Fashion and Design,Donghua University,Shanghai 200051,China;Shanghai Fire Research Institute of MEM,Shanghai 200032,China;Anhui Guqi Down Incorporated Company,Wuhu Anhui 241300,China)

机构地区：[1]安徽工程大学纺织服装学院,安徽芜湖241000 [2]东华大学服装与艺术设计学院,上海200051 [3]应急管理部上海消防研究所,上海200032 [4]安徽古麒绒材股份有限公司,安徽芜湖241300

出　　处：《武汉纺织大学学报》2022年第6期54-58,共5页Journal of Wuhan Textile University

基　　金：安徽省纺织工程技术研究中心和“纺织面料”安徽省高校重点实验室2021年度联合开放基金项目(2021AETKL20);安徽工程大学校级科研项目(Xjky03201908);消防应急救援装备应急管理部重点实验室开放课题(2020XFZB09);2021年省级大学生创新创业训练计划项目(S202110363229);安徽省高等学校自然科学研究项目(KJ2020A0352);2022年安徽工程大学校大学生科研项目(2022DZ18);2022年度安徽工程大学-鸠江区产业协同创新专项基金(2022cyxtb7)。

摘　　要：为解决目前羽毛绒材料定价主要依据经验而缺乏理论支撑和预测精度较低的问题,本文提出利用长短期记忆网络(LSTM)深度学习方法对羽毛绒材料价格进行自定义研究,以2015年-2020年6年的羽绒金网数据为依据,对其构建模型进行训练求解,并与线性自回归移动平均(ARIMA)数理统计模型和最小二乘支持向量机(LS-SVM)浅层机器学习模型预测效果进行对比分析。结果表明:在长期预测中,预测精度从高到低依次为ARIMA模型、LS-SVM模型、LSTM深度学习预测模型;在短期预测中,预测精度从高到低依次为LSTM深度学习预测模型、LS-SVM模型、ARIMA模型。同时还发现:无论长期短期预测中,不同种类的羽毛绒价格预测精度趋势相同,即预测精度仅与资本属性有关,与资本所述类别无关。研究结论既可为羽毛绒企业进行羽毛绒材料准确定价提供理论依据,也为人工智能技术广泛应用于量化投资领域提供实践经验。In order to solve the problem of the forecast-pricing of down material based on experience,due to the lack of theoretical support and low precision of predictive ability,self-definition study of down material price was proposed by the depth learning method of long-term and short-term memory network(LSTM),and was trained and solved based on the data of 6 years from 2015 to 2020,the forecasting results were compared with the linear autoregressive moving average(ARIMA)mathematical statistical model and the Least square support vector machine(LS-SVM)shallow machine learning model,The results show that in the long-term prediction,the order of prediction precision from high to low was ARIMA model,LS-SVM model and LSTM depth learning prediction model;in the short-term prediction,the prediction accuracy from high to low was LSTM depth learning prediction model,LS-SVM model and ARIMA model.And the results showed that the forecast precision of different feather price had the same trend in both long-term and short-term prediction and the forecast precision was independent of capital category instead of capital attribute.The research not only provides theoretical basis for down material pricing,but also provides practical experience for AI technology to be widely used in quantitative investment.

关 键 词：羽毛绒材料 羽绒金网 深度学习 定价模型 预测评测 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]