階層分析法によるデザイン評価データの機械学習を用いた二次元マッピング

抄録

When evaluating a certain design, it is difficult to quantitatively compare the evaluation results because the importance of evaluation criteria differs depending on the evaluator. In this paper, we propose a method that combines the analytic hierarchy process (AHP) and machine learning for the purpose of visualizing design evaluation data. The proposed method generates tree-structured evaluation data based on AHP. Furthermore, the evaluation data are mapped to a low-dimensional space using three machine learning methods. First, the evaluation criteria generated through AHP are converted to feature vectors using a word embedding method. Regarding the tree-structured data in which the evaluation criteria are vectorized as a graph, another feature vectors considering its connectivity are extracted by a graph embedding method. The feature vector associated with the master node is regarded as the whole graph feature, and the vector is mapped to a two-dimensional space using a dimensionality reduction method. The proposed method enables visualizing the similarity of different evaluation data, which provides important feedback to position the evaluators’ decision-making criteria relatively in the data group.