组移动模式挖掘中轨迹聚类的置信区间法

Confidence-interval approach of trajectory clustering for group movement pattern mining of moving objects

在借鉴空间数据挖掘技术的基础上,定义了移动对象轨迹之间的时态距离和平均距离,提出了标准差法和置信区间法两种轨迹聚类算法。两种方法能够找出所有具有相似轨迹的对象对,在不同距离采样点数的基础上配合使用两种方法能够明显降低轨迹聚类算法的时间复杂度。基于标准差法和置信区间法的轨迹聚类算法在仿真数据集和真实数据集进行了验证。表明两种方法能够为其他轨迹聚类算法进行数据筛选,筛选后的数据量将大大减少,从而可提高算法效率。

Based on the spatial data mining algorithms, the temporal distance and average distance of moving objects are defined in this paper, and then sample variance approach and confidence-interval approach for trajectory clustering are provided. The two approaches can discover all the object pairs that have similar trajectories at certain time intervals. Using different sampling granu- larities of trajectory distance can greatly depress the time complexity of the trajectory clustering algorithm. The clustering algo- rithm based on sample variance approach and confidence-interval approach is tested both on synthetic and real datasets. It is indi- cated that the two approaches can also be used as pretreatment methods for other trajectory clustering algorithms, and can greatly reduce the data amount being searched.