基于多线性主成分分析和径向基神经网络的储粮温度变化预测

储粮温度预测有利于及时采取通风降温等措施保障储粮安全,因而具有重要意义。传统预测方法多是基于历史温度数据的向量形式进行特征提取,破坏了原有数据的高阶特性和内部结构,导致局部空间信息丢失,预测精度难以满足要求。针对这一问题,提出了一种以数据张量表示为基础的多线性主成分分析方法。该方法保留了粮堆温度历史数据的高阶性,即在使用多线性主成分分析进行特征提取和降维的同时,充分保留了原有数据的内部结构,因而提取的特征更为有效。试验表明,本方法预测结果优于其他典型预测方法。

A pediction of temperature change tendency for grain storage based on multi-linear principle component analysis and radial basis neural network

A prediction of temperature helps to adopt cooling measures such as ventilate timely to guarantee safety of storage grain.Because the traditional methods destroy high order features and internal structure,which results in the loss of local space information,when abstracting features based on history temperature data,adequate prediction accuracy usually can't be obtained.For the problem,a new method titled multi-linear principle component analysis(MPCA)was proposed based on data tensor representation.This method retained the high order feature of history temperature data of storage grain,to be specific,and retained the internal structure of the original data adequately when abstracting features using MPCA,which made the abstracted features more effective.The experimental results showed that the method was superior to other typical methods.