知识与财富的链接
针对现有质心求解算法仍具有较高计算复杂度,导致区间二型模糊C均值聚类算法(Interval Type-2 Fuzzy C-Means, IT2FCM)运行速度不理想问题,提出了半数迭代法和一次迭代法两种近似质心求解算法。首先,在直接求解转换点问题质心求解算法(A Direct Approach for Determining the Switch Points in the Karnik–Mendel Algorithm, DA)的基础上,借助二分查找思想,构造出基于二分查找的质心求解算法;接着,以该算法为基础,通过限制查找范围,考虑两个转换点之间关系的性质和计算差值的技巧得到半数迭代法;最后,考虑只进行一次查找得到一次迭代法。在UCI上的5个数据集上(IRIS、SEEDS、WINE、WIFI＿LOCALIZATION和HTRU2)验证了两种算法的聚类性能并没有因为求解的是近似质心而降低;进一步在ANURAN CALLS数据集上构造了8组数据量递增数据用于验证基于不同质心求解算法的IT2FCM和基于提出的近似质心求解算法的IT2FCM运行速度,实验结果表明：基于近似质心求解算法的IT2...
In view of the high computational complexity of the existing centroid solving algorithms leading to the unsatisfactory running speed of interval type-2 fuzzy C-means (IT2FCM) clustering algorithm, two approximate centroid solving algorithms, including half iterative method and one-time iterative method, were proposed. Firstly, based on a direct approach for determining the switch points in the Karnik-Mendel algorithm, with the help of the idea of binary search, a centroid solving algorithm based on binary search was constructed. Based on this algorithm, by limiting the search range, and considering the nature of the relationship between the two conversion points and the skill of calculating the difference, the half iterative method was obtained. Finally, the one-time iterative method was obtained by considering only one search. Five data sets (IRIS, SEEDS, WINE, WIFI_LOCALIZATION and HTRU2) on UCI verified that the clustering performances of the two algorithms were not reduced by solving approximate centroid. Further, eight groups of incremental data were constructed on ANURAN CALLS data set to compare the running speed of IT2FCM based on different centroid solving algorithms and IT2FCM based on the proposed approximate centroid solving algorithm. The experimental results showed that IT2FCM based on approximate centroid solving algorithm ran faster. Therefore, the proposed approximate centroid algorithm can alleviate the high complexity of IT2FCM to a certain extent.