一种基于信息熵的时空点模式分析方法

A Method of Spatio-Temporal Point Pattern Analysis Based on Information Entropy

现有时空点模式分析方法在度量时空邻近或时空密度时,存在时空耦合参数选择的主观性问题,无法得到有效的分析结果,为此,该文提出了一种基于信息熵的时空点模式分析方法。首先,计算每个时空点实体的空间最近邻的时间距离;进而,统计不同范围内空间最近邻的时间距离的频率分布特征,计算归一化信息熵值描述分布的不确定性程度,归一化熵值越大越表现为聚集分布,熵值越小越趋近于随机分布。最后进行了模拟实验比较和实际应用验证分析,结果表明:该方法在无须输入敏感性参数条件下,能够识别不同点模式类型,并能近似度量不同强度的聚集模式。

Existing methods of spaaior-temporal point pattern analysis require users to set both space and time autocorrelation parameters, and the selection of those parameters is difficult and suljective, which makes it difficult to obtain effective and reliable results.To overcome this difficulty,in this paper, a spatio- temporal point pattern analysis method based on the information entropy theory is proposed. In the method,the temporal distance of one object to its spatial nearest neighbor is firstly calculated, and then the frequency of the temporal distance is counted in each interval. Finally, an index based the normalized information entropy is developed to describe the uncertainty of the temporal distance distribution. According to the value of the index, the type of spatio-temporal point pattern can be identified, and a larger value indicates an aggregate distribution pattern while a small value （which is near zero） means a random distribution pattern. Both simulated and rea-life datasets are used to evaluate the proposed method, and the results show that the proposed method can identify the different point pattern types with less parrameters, and what s more, it has been found that the index can even indicate the strength of aggregation patterns in some sense.