中文摘要：

为准确推断时间序列间的因果网络,针对传统因果强度衡量方法的不足,提出一种基于信息熵的因果强度衡量标准（归一化因果熵）。为改进传统方法量纲不统一且冗余较多的缺点,通过归一化处理使量纲不同的节点间强度具备可比性,通过排除节点间的间接影响大幅减少冗余,更准确地衡量时序节点间的因果强度;在此基础上,设计时间序列的因果推断算法,以归一化因果熵衡量节点间因果关系的强弱,筛选强关系形成完整因果图。实验结果表明,该算法相比起传统算法更准确有效。

英文摘要：

To infer the causal network of time series accurately, a measure of causal strength based on information entropy （nor- malized causal entropy） was proposed to overcome shortcomings of traditional measure of causal strength. To improve the traditional measure, which does not uniform dimension, causing much redundancy, normalization was used to make strength between nodes with different dimension comparable, while significantly reducing redundancy by excluding indirect effect between nodes. Normalized causal entropy measure causal strength between time series was then more accurate. Based on these, a time-series causal network inference algorithm was designed, and the normalized causal entropy was used as the measure of causal strength between nodes, screening the strong relationship to complete the causal map. Experimental results show that the algorithm is more accurate and effective compared to traditional algorithms.