Using topology in data analysis, known as Topological Data Analysis (TDA), is now a promising new area of data mining research. One of the important and foundational tools of TDA is the Mapper algorithm. During the past two decades, this algorithm has proven its useful and robust abilities in extracting insights and meaningful information from high-dimensional datasets. Nevertheless, several alterations in the choices of parameters, such as lens, cover and clustering, can be used to develop this algorithm. In this paper, we propose the F-Mapper algorithm, based on the foundation of the Mapper algorithm, to solve the problem of automating when dividing cover intervals with an arbitrary percentage of overlap. To clarify the efficiency of this enhanced algorithm, experiments were carried out on three datasets, including the Unit Circle, Reaven and Miller Diabetes, and NKI Breast Cancer. The experimental results will be analyzed and compared with those of the original method, the Mapper algorithm, through the output image and silhouette coefficient score in the evaluation of clustering.

在数据分析中使用拓扑，即拓扑数据分析（TDA），现在是数据挖掘研究中一个有希望的新领域。TDA的重要基础工具之一是Mapper算法。在过去的二十年中，该算法已证明其从高维数据集中提取见解和有意义的信息的有用和强大的功能。不过，可以使用参数选择的几种更改来开发此算法，例如镜头，覆盖物和聚类。在本文中，我们基于Mapper算法的基础，提出了F-Mapper算法，以解决以任意百分比的重叠率划分覆盖区间时的自动化问题。为了阐明此增强算法的效率，我们在三个数据集中进行了实验，包括单位圆，Reaven和Miller糖尿病以及NKI乳腺癌。通过对输出图像和轮廓系数得分进行聚类评估，将实验结果与原始方法Mapper算法进行分析并进行比较。