CAJ | 학술논문

确定时间序列的相似性匹配方法都没有考虑数据的不确定性，而现实世界中诸如温度传感器等设备采集到的数据往往是不确定的，并且两条不确定时间序列之间的距离也是不确定的，所以现有的确定时间序列的相似性匹配方法不适用于这些领域。针对此问题，提出了基于统计学的规约算法，并且基于该算法提出了不确定时间序列相似性匹配的两种新型算法。在规约过程中，规约算法优化了不同背景下不确定时间序列的小概率点和奇异点的处理。在匹配过程中，首先提出了圆环匹配算法，它通过构建匹配圆环完成相似性匹配，并且通过多次重启提高相似性匹配的准确度和效率；然后在规约算法的基础上，提出了期望匹配的改进算法，它通过增加包络约束消除期望匹配算法中出现的误判问题。
학정시간서렬적상사성필배방법도몰유고필수거적불학정성，이현실세계중제여온도전감기등설비채집도적수거왕왕시불학정적，병차량조불학정시간서렬지간적거리야시불학정적，소이현유적학정시간서렬적상사성필배방법불괄용우저사영역。침대차문제，제출료기우통계학적규약산법，병차기우해산법제출료불학정시간서렬상사성필배적량충신형산법。재규약과정중，규약산법우화료불동배경하불학정시간서렬적소개솔점화기이점적처리。재필배과정중，수선제출료원배필배산법，타통과구건필배원배완성상사성필배，병차통과다차중계제고상사성필배적준학도화효솔；연후재규약산법적기출상，제출료기망필배적개진산법，타통과증가포락약속소제기망필배산법중출현적오판문제。
Similarity matching techniques for certain time series did not consider the uncertainty of the data,but the data col-lected by the sensors were often not certain in the real world.So,the existed similarity matching methods of time series did not apply to these areas.To solve this problem,this paper put forward a reduction algorithm based on statistics and improved the Euclidean distance calculation.Then,it raised new similarity matching algorithms.In the process of reduction,it optimized the treatment of small probability points and singular points in every slot.In the process of circle similarity matching,it im-proved the accuracy and decreased the time cost by restarting many times.At last,it put forward the improved algorithm by ex-pectancy calculation based on reduction algorithm and solved the miscarriage of justice problem.