現今，決策樹分類法要求屬性及類別標籤均須為單值。然而，真實世界存在著多值多標籤的資料，為了能處理此種多值多標籤資料的分類，本研究首先設計了一個決策樹分類法並命名為MMC (Multi-valued and Multi-labeled Classifier)；其次，藉由重新設計此演算法，我們發展另一個分類法並命名為MMDT (Multi-valued and Multi-labeled Decision Tree) 以改善 MMC 的正確率。 MMC 和 MMDT不同於傳統決策樹分類法的一些主要功能，包括生長決策樹、選擇屬性、以標籤代表葉節點及預測新的資料。MMC的發展策略主要基於多標籤間的相似度測量，而MMDT 的發展策略主要暨基於多標籤間的相似度測量及評分。 實驗結果說明 MMC 和 MMDT 不僅能從大量的多值及多標籤資料集來挖掘出規則，而且得到具說服性的正確率和規則良好度。 Presently, decision tree classifiers require that attributes and class label of data set to be single-valued. However, there exist classification problems with multi-valued and multi-labeled data. Aiming to handle this multi-valued and multi-labeled data, this research has developed a decision tree classifier named MMC (Multi-valued and Multi-labeled Classifier) first. Then, by redesigning the algorithm, this research has further developed another classifier named MMDT (Multi-valued and Multi-labeled Decision Tree) to improve the accuracy of MMC. MMC and MMDT are different from the traditional decision tree classifiers in some major functions including growing a decision tree, selecting attribute, assigning labels to represent a leaf and making a prediction for a new data. The development strategy of MMC is mainly based on measuring similarity among multiple labels; the development strategy of MMDT is mainly based on both measuring similarity and scoring among multiple labels. The experimental results show that MMC and MMDT can not only mine classification rules from a large multi-valued and multi-labeled data set, but also get convincing accuracy and goodness of rules.
