基于有向有限穿越可视图的时间序列伴生网络

摘要

提出两种新的可视图建网方法(有限穿越水平可视图和有向有限穿越可视图),有限穿越水平可视图综合了水平可视图的算法简便性以及有限穿越可视图的抗噪声能力,而有向有限穿越可视图则突破了原有算法的无向性,考虑了时间的不可逆性.与传统可视图的三种模型进行对比,验证了两种算法的有效性,最后将模型应用到股市进行了实证分析.%This paper proposes two models:limited penetrable horizontal visibility graph(LPHVG) and directed limited penetrable visibility graph(DLPVG).LPHVG has both the simplicity of horizontal visibility graph(HVG) and the anti-noise performance of LPHVG.Directed limited penetrable visibility graph(DLPVG) takes into account the irreversibility of time and makes a breakthrough in the non-directional algorithms.The comparative experiments verifies the effectiveness of the two models,and the models are applied empirically to the real stock markets.