How to Control Clustering Results? Flexible Clustering Aggregation

Evolving Ensemble-Clustering to a Feedback-Driven Process

“…In this section, we introduce our ensemble clustering approach-Flexible Clustering Aggregation (FCA) [3]. This approach represents the algorithmic platform of our feedback-driven ensemble-clustering process.…”

“…The basic concept of FCA is clustering aggregation, which combines different clusterings of a dataset into one result to increase quality and robustness [3], [4]. Different aggregation approaches are known, where the pairwise assignment approach is considered as the most capable one.…”

“…Despite the multitude of available algorithms, data clustering remains a very challenging task, especially from the end user's perspective. The core problem is to select an optimal algorithm and parameterization for the particular task at hand, whereas it is not guaranteed that either of them exists [1], [2], [3]. So, in practice, the user has to complete the three steps: (i) algorithm selection, (ii) algorithm parameterization and execution, and (iii) result interpretation.…”

“…Subsequently, the employed algorithmic platform is introduced in Sec. III, which is based on our ensemble-clustering approach [3]. Using this platform as a foundation, we define the necessary components of our process in Sec.…”

“…Using this platform as a foundation, we define the necessary components of our process in Sec. IV: First, we derive the feedback necessary to navigate through our process from our aggregation method [3], which features an intuitive effect-oriented parameterization(Sec. IV-A).…”