论文部分内容阅读
将M-估计稳健损失函数与基于统计贡献度的动态确定核函数方法相结合,提出一种有效的非参数RBF预测模型,该方法克服了稳健性缺失问题,在估计参数的同时动态确定最佳网络结构,并且在学习中自动消除噪声和异常点的影响,加快了网络的学习和收敛速度.利用中国月度信贷数据进行实证分析表明,本文模型与基准模型相比具有最好的预测稳健性和准确性,对于提高货币政策有效性和前瞻性具有很好的应用价值.
Combining the M-estimator robust loss function with the dynamic deterministic kernel function method based on statistical contribution, an effective non-parametric RBF prediction model is proposed, which overcomes the problem of lack of robustness and dynamically determines the optimal Network structure and automatically eliminate the influence of noise and abnormal points in learning to speed up the network learning and convergence rate.Using the Chinese monthly credit data for empirical analysis shows that the model compared with the benchmark model has the best prediction robustness and Accuracy, for improving the effectiveness of monetary policy and forward-looking has good application value.