【摘要】 金融时间序列一直以来以其非线性、非平稳、信噪比低等特性成为时间序列预测中的难题.本文提出基于变分模态分解的VMD-ELM模型,利用变分模态分解在复杂的金融时间序列数据分解上的特有优势,将金融时间序列数据分解为若干个子模态,再将分解后的子模态作为极限学习机的输入数据进行训练.基于平均绝对误差(MAE),平均绝对百分比误差(MAPE)和平方根均方根(RMSE),通过比较EMD-ELM模型,前馈神经网络(FFNN)和自回归移动平均(ARMA)在西德克萨斯中质原油(WTI),加拿大/美国汇率(CANUS),美国工业生产(IP)和芝加哥期权交易所纳斯达克100波动率指数(VIX)时间序列数据上的效果,表明本文提出的方法在多个数据集上均有优秀的表现.更多还原

【Abstract】 Financial time series has always been a problem in time series forecasting due to its non-linear,non-stationary,and low signal-to-noise ratio. This paper proposes a VMD-ELM model based on the variational modal decomposition,which uses the unique advantages of variational modal decomposition in the decomposition of complex financial time series data to decompose financial time series data into several sub-modes,and then the decomposed sub-modes are used as the input data of the extreme learning machine for training. By comparing the different MAE,MAPE and RMSE results of the EMD-ELM,FFNN and ARMA on the WTI,CANUS,IP and VIX dataset,it is proved that the method proposed in this paper has excellent performance on multiple datasets.更多还原