Non-parameter clustering algorithm based on saturated neighborhood graph

Clustering algorithms play a very important role in the field of data mining and machine learning. However existing clustering methods are sensitive to parameters and outliers. The commonly used clustering methods are restricted by problem of parameter selection that different algorithms need to input one or more different parameters. For overcoming these drawbacks, we propose a non-parameter clustering algorithm based on saturated neighborhood graph known as NPCSNG. NPCSNG algorithm uses mathematic method to preprocess the data set, and then uses the characteristics of SNG adaptive clustering to cluster the data, so as to achieve the purpose of non-parameter clustering. NPCSNG has three main advantages: (1) it does not need to manually set any parameters due to the use of adaptive saturated neighborhood graph; (2) it significantly improves the clustering performance as well as the model robustness, making NPCSNG a more practical approach for real-world scenarios; (3) it can easily adapt to data-sets with complex manifold structure. NPCSNG algorithm solves the problem of parameter selection of clustering algorithm and it broadens the idea of clustering by using the characteristics of graphs. (c) 2022 Elsevier B.V. All rights reserved.

Automated summary

This paper introduces a non-parameter clustering algorithm based on saturated neighborhood graph (NPCSNG), which preprocesses the data set using mathematical methods and clusters the data using the characteristics of SNG adaptive clustering to achieve non-parameter clustering. The NPCSNG algorithm has the advantages of not requiring manual parameter setting, significantly improving clustering performance and model robustness, and adapting easily to data sets with complex manifold structure.