CAJ | 학술논문

传统 FCM 聚类算法存在初始聚类中心较为敏感的问题，易陷入局部最优。针对此问题，提出了基于密度权值和自适应免疫系统的 FCM 算法（ d-AIFCM ）。算法在对 Web 用户进行聚类分析的过程中，基于用户群体之间的相似性，引入密度权值生成候选初始聚类中心，采用自适应免疫系统的原理确定初始聚类中心，自动生成最佳分类，解决传统FCM 算法对初始聚类中心敏感的问题。实验结果表明， d-AIFCM 算法在收敛次数和聚类效果方面较其他同类算法有所提升。
전통 FCM 취류산법존재초시취류중심교위민감적문제，역함입국부최우。침대차문제，제출료기우밀도권치화자괄응면역계통적 FCM 산법（ d-AIFCM ）。산법재대 Web 용호진행취류분석적과정중，기우용호군체지간적상사성，인입밀도권치생성후선초시취류중심，채용자괄응면역계통적원리학정초시취류중심，자동생성최가분류，해결전통FCM 산법대초시취류중심민감적문제。실험결과표명， d-AIFCM 산법재수렴차수화취류효과방면교기타동류산법유소제승。
The traditional FCM clustering algorithm exists the problem that the initial clustering center is relatively sensitive , which easily plunges into local optimum. Aiming at this problem, this paper proposes the d-AIFCM algorithm which is based on the density of weights and the adaptive immune system. According to the similarity between user groups, when this algorithm carries out clustering analysis on the Web users, it introduces density weights to generate candidate initial clustering center, uses the principle of the adaptive immune system to determine the initial clustering center, automatically generates the best classification, and solves the problem that the traditional FCM algorithm is sensitive to the initial clustering center. The experiment results show that d-AIFCM algorithm is much more improved than other similar algorithm in terms of convergence times and clustering effect.