在科技不斷進步的時代下，許多問題是越來越複雜，分類任務也是如此，為了處理更貼近生活的複雜問題，以電影為例，一部喜劇電影所涵蓋的元素可能不單僅有喜劇元素，而是涵蓋了動作、喜劇與愛情等的元素，然而傳統分類模型只能歸類於單一類別，不符合現實生活的景況，因此就需要以多標籤分類方式來協助人們分類。多標籤分類方法有兩大派系，分別為問題轉換與改編算法，兩種方式在多標籤分類方法中有許多不同的演算法，而本研究透過問題轉換與改編算法兩大派系獲得靈感。本研究主要以關聯規則找出屬性值與屬性值之間的關係，將具有強烈關係的屬性值轉換為屬性與屬性之間的關係，並將它們合併，改變資料集結構，再加入至學習模型中加以訓練，目的在於提升模型精準。最後，在研究方法中，會加以研究關聯的種類、分類與其特性，並發展以關聯規則為基礎之多標籤分類演算法，而關聯規則的好處在於任何屬性之間都能存在關聯，它會嘗試尋找許多規則，每個規則可能具有不同之結果。

Science and technology change rapidly in this age. Problems are becoming more and more complicated and so do Multi-label classification problems. To deal with the complicated problems in our life, multi-label classification is what we need. There are two major factions of the Multi-label classification method, called “Problem Transformation” and “Algorithm Adaption Method.” There are many algorithms in both major factions, and our research got inspiration from both of them. Moreover, we will utilize five evaluation measures in the future experiment, like “Hamming Loss”, “One-Error”, “Coverage”, “Ranking Loss”, and Average Precision. Our research is to find a strong correlation between attribute values. And we will transform the correlation between attribute value to correlation between attribute. Then, merge the attributes which have strong correlation to alter the dataset. Furthermore, add it to the learning model for training to improve the accuracy of the model. Finally, we will study the type, classification and characteristics of the association and develop the association analysis algorithm based on association rules in the future experiment.