基于CNN-LSTM-STL多任务模型的陶瓷艺术品价格预测研究  

Research on Ceramic Art Price Prediction Based on CNN-LSTM-STL Multi-Task Model

作　　者：徐永欢 詹棠森 XU Yonghuan;ZHAN Tangsen(School of Information Engineering,Jingdezhen Ceramic University,333403,Jingdezhen,Jiangxi,PRC)

机构地区：[1]景德镇陶瓷大学信息工程学院,江西景德镇333403

出　　处：《江西科学》2024年第6期1301-1305,共5页Jiangxi Science

基　　金：国家自然科学基金项目(71763013)。

摘　　要：随着中国经济的高速发展,我国的陶瓷艺术品市场也迅速发展起来,但在价格上却并没有形成系统的制度,造成了价格上的差异。基于雅昌拍卖网的150组陶瓷数据,先利用卷积神经网络(CNN)进行特征提取,将提取得到的数据输入LSTM模型中,最后将CNN-LSTM模型作为多任务模型中的一个单任务模型,从而得到了改进的CNN-LSTM-STL多任务预测模型。经过实证说明,改进CNN-LSTM-STL预测模型显著提高了序列的模拟和预测的精度,其研究结果为陶瓷艺术品价格的预测提供了依据。With the rapid development of the China’s economy,the ceramic art market inChina has also rapidly developed.However,a systematic pricing framework has not yet beenestablished,leading to significant price disparities.This paper utilizes 150 sets of ceramicdata from Yachang Auction website.First,convolutional neural networks(CNN)are usedfor feature extraction,and the extracted data is input into the LSTM model.Finally,theCNN-LSTM model is used as a single task model in the multi-task model,resulting in animproved CNN-LSTM-STL multi-task prediction model.Empirical results show thatthe improved CNN-LSTM-STL prediction model significantly enhances the accuracy ofsequence simulation and prediction,and its research results provide a basis for predicting theprice of ceramic artworks.

关 键 词：陶瓷艺术品 卷积神经网络(CNN) 多任务预测 CNN-LSTM-STL预测模型 

分 类 号：TQ174.74[化学工程—陶瓷工业]