ツリー分割によるアノテーションの可視化法

DOI Web Site
  • 小林 えり
    静岡県立大学経営情報イノベーション研究科
  • 伏見 卓恭
    静岡県立大学経営情報イノベーション研究科
  • 斉藤 和巳
    静岡県立大学経営情報イノベーション研究科
  • 池田 哲夫
    静岡県立大学経営情報イノベーション研究科

書誌事項

タイトル別名
  • Visualization of annotation by tree division
  • ツリー ブンカツ ニ ヨル アノテーション ノ カシカホウ

この論文をさがす

抄録

<p>When analyzing visualization results, we want to grasp where such objects as have a certain type of features are located in. To this end, we propose a novel technique which automatically divides all objects into some subsets each of which has the characteristic feature distribution substantially different from that of other subsets, and annotates the characteristic feature to each subset. Concretely, our proposed method includes three steps: first, constructs a minimum spanning tree based on distances between each object; second, cuts the tree edge where the feature distributions substantially change, and divides into some subsets of objects; finally, extracts characteristic features using Z-score measure. In our experiments using two datasets including hand-written characters and restaurant location coordinates, we confirmed that our method can automatically detect the subsets which have characteristics feature distributions, and may contribute to the improvement of readability of visualization results.</p>

収録刊行物

詳細情報 詳細情報について

問題の指摘

ページトップへ