D BCMO -Mapper: A Metaheuristic Framework for Distribution-Guided Parameter Optimization in Topological Data Analysis
Mapper is a widely used tool in Topological Data Analysis (TDA). It represents complex data as graphs, helping users identify meaningful structures. However, one of the factors determining the quality of results obtained using the Mapper algorithm is the input parameters. This leads to instability and limits the reproducibility of results. The Distribution-Guided Mapper (D-Mapper) was developed to overcome this limitation. By using a Gaussian Mixture Model (GMM) to adjust the width of intervals based on the data distribution density, D-Mapper has partially overcome the limitations of Mapper. However, D-Mapper still depends on user-specified parameters, so it has not yet solved the core problem of Mapper. To address the remaining dependence on user-specified parameters, we propose DBCMO-Mapper as an automated parameter-selection extension of D-Mapper. Balancing Composite Motion Optimization (BCMO) searches for the number of GMM components and the confidence coefficient within predefined bounds using the adjusted silhouette measure SC adj as its objective. Experiments on synthetic and real-world datasets illustrate the Mapper representations obtained under automatically selected parameters. The framework provides a systematic parameter-selection procedure while retaining the distribution-guided covering and graph-construction procedure of D-Mapper.
Authors
- Hung Thanh Nguyen (ORCID: https://orcid.org/0000-0001-6770-0528)
- Minh-Huy Bui (ORCID: https://orcid.org/0009-0007-1550-7970)
- Loi Ha
- Quang-Thinh Bui
Publication Details
- Journal
- Vietnam Journal of Computer Science
- Published
- 2026-11-13
- DOI
- https://doi.org/10.1142/s2196888826500223
- Primary Topic
- Topological and Geometric Data Analysis
- Type
- article
- Field-Weighted Citation Impact
- 0.00