金融时序数据具有非平稳、含噪声等特点,采用单一的预测模型时预测精度并不理想.据此,构建了基于小波去噪的灰色和支持向量机组合预测模型,并对预测难度较高的我国热钱流动规模进行了预测.实验结果表明:与经典GM(1,1)幂模型相比,组合预测模型的平均相对误差由46.14%降到2.95%,预测精度有显著性提高. The financial time series data are non-stationary and noisy, and the prediction accuracy when using a single prediction model is not ideal.Accordingly, a combined forecasting model of gray and support vector machine based on wavelet denoising is constructed, and the prediction model with high prediction difficulty The results show that compared with the classical GM (1,1) power model, the average relative error of the combined forecasting model is reduced from 46.14% to 2.95%, and the prediction accuracy is significantly improved.