【摘要】 预测模型中的样本存在一定的误差和孤立点,这种现象普遍存在于控制领域的预测过程中。而样本学习过程往往呈现出增量性、批量性和在线性3大特性,为了保证模型预测的精度并且使预测模型可以随着样本的增加而动态的更新,本文提出了一种基于支持向量机回归的批量样本的增量式在线学习算法(B IO)。算法通过对矫直钢板平直度进行预测并对建模的工业实例进行研究。结果表明:本算法不仅提高了模型的预测精度并且使模型具有实时在线更新的能力。更多还原

【Abstract】 There are always some kinds of error and isolated samples in the training samples in the predicting model.The learning process of samples always presents three major characteristics: batch,increment and online.We propose a learning algorithm of batch,increment and online(BIO) based on support vector regression which can ensure the accuracy of the predicting model and update dynamically when the samples increase.When it is used in industry,our algorithm can analyze and predict the flatness of plate.Result shows that compared with the traditional incremental support vector machine(SVM),our algorithm not only improves the accuracy of the model but also make the model capable of real-time updating.更多还原