On Learning Classifier System Clustering And Backbone Extraction Methods Under Unsupervised Learning Framework

Unsupervised learning is one of the most important research branch in the field of machine learning and its application is very extensive. For example: clustering of data extraction, complex network backbone etc. Based on the voting integrated clustering and complex cluster network diagram as a starting point for research, we obtain the following results:(1) To solve data clustering tasks, a novel consensus clustering method based on extended classifier system is proposed, called voting-XCSc. When conducting the clustering for the data points, the proposed method will first employ the XCSc to generate a set of clustering results with different clustering numbers, and then it will adopt the dissociation-based strategy to experimentally determine the clustering number among all the candidates. Finally, a majority voting-based consensus method is applied to obtain the final clustering results. The proposed voting-XCSc has been evaluated on both the toy examples as well as two real clustering-related applications.(2) To solve data clustering tasks, a novel clustering ensemble-based framework based on extended classifier system is proposed. The framework includes more integration guidelines, consensus function and adaptive integration and so on. Specifically, when dealing with clustering task, our proposed methods will first generate several base clustering results by performing XCSc, with the goal of keeping the large diversity by clustering the data with different initialization(e.g., number of clusters). With obtaining the base clustering results, the proposed methods will then employ the corresponding strategies to generate the final clustering result. We systematically evaluate the proposed methods on both the synthesis example as well as two real clustering-related applications. The experimental results validated that by different base clustering results ensemble, the results can be improved compared with performing XCSc independently.(3) To solve the graph clustering tasks in complex network, a backbone extraction heuristic with incomplete information(BEHw II) is proposed to find the backbone in a complex weighted network. The backbone is the natural abstraction of a complex network, which can help people understand a networked system in a more simplified form. In our method, a novel edge-filtering rule is first designed based on the null model. Thus, we present a local search model to examine part of edges in an iterative way. Experimental results on four real-life networks demonstrate the advantage of BEHw II over the classic disparity filter method by either effectiveness or efficiency validity.