论文部分内容阅读
针对集成在线序贯极端学习机(EOS-ELM)预测精度不高和动态适应性差的问题,提出一种具有选择与补偿机制的加权集合序贯极端学习机.该加权集合序贯极端学习机在序贯学习过程中,通过对当前预测模型精度的判断决定是否进行递推更新操作,同时为提高预测模型的动态跟踪能力,在加入新样本的同时对旧样本进行剔除;然后,利用EMD对残差序列处理后进行预测,并将初始预测结果与残差预测结果相加得到最终预测模型.通过对上证指数的预测,结果表明所提方法具有更好的泛化性能,预测精度相比EOS-ELM提高了近36.1%.
Aiming at the problem of poor accuracy and poor dynamic adaptability of integrated online sequential extreme learning machine (EOS-ELM), a weighted set sequential extreme learning machine with selection and compensation mechanism is proposed. The weighted set sequential extreme learning machine In the process of sequential learning, the accuracy of the current prediction model is used to decide whether to perform the recursive update operation. In order to improve the dynamic tracking ability of the prediction model, new samples are removed while the new samples are added. Then, After the sequence is processed, the prediction is made, and the initial prediction result and the residual prediction result are added to get the final prediction model.Through the prediction of the Shanghai Composite Index, the results show that the proposed method has better generalization performance, the prediction accuracy is better than EOS- ELM increased by nearly 36.1%.