CAJ | 학술논문

K‐均值聚类算法因具有实现简单、快速有效、适合处理大数据集等优点而被广泛应用。但是由于初始聚类中心是随机选择的，K‐均值聚类的结果对初始中心值具有很大的依赖性。另一方面，聚类的类数 K 也对聚类结果具有重要影响，一般情况下 K 是未知的，需要相关的专业知识对 K 进行预测，如果不能事先确定合适的 K 值，聚类的结果也会很不理想。本文提出一种改进的 K‐均值算法，避开了初始中心点的随机选择，只需计算数据集合中相距最远的两个点将其设为初始中心，同时不需要预先猜测聚类的数目，大大减小了用户的使用难度，聚类效果也显著提高。论文将改进的 K‐均值算法应用到自然雷电脉冲的分类中，并与传统 K‐均值算法的分类结果进行了比较。
K‐균치취류산법인구유실현간단、쾌속유효、괄합처리대수거집등우점이피엄범응용。단시유우초시취류중심시수궤선택적，K‐균치취류적결과대초시중심치구유흔대적의뢰성。령일방면，취류적류수 K 야대취류결과구유중요영향，일반정황하 K 시미지적，수요상관적전업지식대 K 진행예측，여과불능사선학정합괄적 K 치，취류적결과야회흔불이상。본문제출일충개진적 K‐균치산법，피개료초시중심점적수궤선택，지수계산수거집합중상거최원적량개점장기설위초시중심，동시불수요예선시측취류적수목，대대감소료용호적사용난도，취류효과야현저제고。논문장개진적 K‐균치산법응용도자연뇌전맥충적분류중，병여전통 K‐균치산법적분류결과진행료비교。
K‐means has gained popularity because of its simplicity and rapid speed of classifying massive data rapidly and efficiently .However ,the output of K‐Means clustering algorithm highly depends upon the selection of initial cluster cen‐ters because the initial cluster centers are chosen randomly .The other limitation ofthe algorithm is to input the required num‐ber of clusters .This requires some sort of intuitive knowledge about appropriate value of K which is sometimes difficult to predict as it requires domainknowledge .If the value of K is not appropriate ,the output of K‐Means clustering algorithm will be bad .In this paper ,we have proposed an algorithm based on the K‐Means ,but it avoids randomly choosing of the initial cluster centers ,only setting the farthest two points in the data set as theinitial cluster centers .On the other hand ,it does not require the number of clusters K as input .It greatly reduces the user's difficulty and increases the quality of the result .This paper applys the improved K‐means clustering algorithm to the classification of natural lightning pulses and makes a compari‐son with traditional K‐means clustering algorithm .