「資料探勘」的定義就是從大型資料庫中發現知識，將隱含的、先前並不知道的、潛在有用的資訊從資料庫中粹取出來的過程(楊昇宏，2000)。在大量的音樂資訊中，如何快速找到使用者會喜歡的音樂？為了解決新資源或是新使用者的冷啟動（Cold-start）問題，本研究利用「內容推薦」以及「資料探勘」中「決策樹」、「貝式網路」，及「類神經網路」的技術。不管是新曲目或是原來的曲目，從預測使用者對新曲目的喜好開始，僅需少量的使用者資料，就可以推薦一首使用者可能沒聽過，但是會喜歡的曲目。利用「內容推薦」及「資料探勘」的特性，更能有效達到預測的目的。
The distribution of sound has been existent for dozens of years. Even nowadays, radio broadcast still plays an indispensable role in our daily lives. With the prevalence of listening to music or downloading music on the Internet, radio broadcast has existed in the form of an Internet medium. Internet radio stations broadcast sound and music to the world. With the aid of a music recommendation system, how should they pick music from a large database and according to the historic preference of users and recommend it?
Data mining is a process of discovering knowledge from large databases and
ex-tracting underlying, unknown, and potentially useful information from these data-bases (Yang, San-Hung, 2000). How to efficiently locate the music that users prefer among the tremendous music databases? For solving the problems of "new resources" and "cold start for new users", this research used theories of Content-Based Filtering, Data Mining (including Decision Tree, Bayesian Network and Neural Networks). Whether new or old materials, starting from the prediction the music preference from new users, and it takes only few user information to recommend a song that user never heard but may appreciate. Utilize the specifics of " Content-Based Filtering " and "Data Mining", can efficiently reach the goal of prediction.
