An Improved Three-Way Clustering Based on Ensemble Strategy

As a powerful data analysis technique, clustering plays an important role in mining. Traditional hard uses one set with crisp boundary to represent cluster, which cannot solve the problem of inaccurate decision-making caused by information or insufficient data. In order this problem, three-way was presented show uncertainty dataset adding concept fringe region. paper, we present improved algorithm based on ensemble strategy. Different existing methods using various algorithms produce base results, proposed randomly extracts feature subset samples and traditional obtain diverse results. Based labels matching is used align all results given voting method core region three way clustering. The can be applied top any generate examples for demonstration, apply K-means spectral clustering, respectively. experimental that effective revealing cluster structures.