基于地理位置时间序列的相似性研究

Research on similarity based on location time series

位置时间序列常被应用到不同的研究领域,具有很高的商业价值。本文通过计算位置时间序列的余弦相似度找到确定与其近似的位置时间序列。同时针对余弦相似度在计算位置时间序列相似性出现的偏差,提出了一种余弦相似度的改进方法(单侧相似度)。单侧相似度给出了不同位置时间序列的包含关系,这种关系是余弦相似度的进一步解释,可以用来衡量不同研究对象之间的包含程度。通过实验,证明了单侧相似度更适合描述位置时间序列的相似性。

Position time series are often used in different research fields, and have high commercial value. By calculating and comparing the similarity of position time series, we can find the approximate position time series. For the deviation appears in similarity calculation of position time series by cosine similarity, an improved method（unilateral similarity） was proposed based on cosine similarity. Unilateral similarity given a containment relationship between different position time series, this relationship is further explained cosine similarity, which is used to measure the containment relationship between different research object. The experimental results show that the unilateral similarity is more suitable for describing the similarity of position time series.