群組飛輪音樂推薦系統之研究與實作

音樂在飛輪運動中屬於非常重要的一環，飛輪課程在教練與音樂的帶動下可以產生多種動作的配合來達到有氧與無氧之間的間歇運動，而如何滿足在同一堂課程中大部分學員的音樂偏好就顯得非常重要，因此研究在飛輪運動中使用群體音樂推薦是否可行是非常實用的議題。　　本研究使用模糊評分與影響權重的方式導入現有的群體音樂推薦方法，並透過飛輪運動比較在未使用此方法選擇音樂時與使用此方法推薦音樂時的評分變化。為了提升學員對飛輪課程音樂的共鳴與降低教練選擇音樂的負擔，因此加入群體音樂推薦，而音樂的評分全數採用模糊評分的方式，並使用影響權重的方式調整學員之間對音樂的偏好關係，在推薦的過程中運用決策樹來模擬每位學員對音樂的偏好，進而得出那些未被聽過但有潛力的音樂之評分。　　為此我們設計了一套推薦系統，除了能幫助教練安排適當的音樂或歌曲，且這些音樂或歌曲也能同時滿足學員們的喜好。學員喜好的音樂不僅能讓學員更加地融入到課程中，也會因為聽到喜好的音樂進而刺激學員達到更好的運動效果。根據實驗結果，學員對於系統所推薦的音樂滿意度也有逐漸上升的情況，滿意度從61.82%提升至89.09%，證明此系統能達到此研究的目標。

關鍵字

群組推薦系統；決策樹；飛輪運動；模糊評分

並列摘要

Music plays an important role in Spinning. During lessons, leading by the coach and music can bring out various of movement combinations to achieve interval exercise in aerobic and anaerobic exercise. Therefore, how to get a music that most of the students in same lesson preference becomes important. So doing research for the reliability of group music recommendation in spinning is a practical issue. This research implements fuzzy score and influence weight to group music recommendation and combine with spinning and compare the score changing to before and after the method apply. In order to provoke more sense of emotional resonance in music amongst the learners and to decrease the coach’s pressure of selecting music ,we implement group music recommendation which uses fuzzy score method to score music and adjust the music preference between different learners by influence weight method. During recommending, we also have decision tree analysis method to simulate every learner’s preferences in music so that we can score some potential music that haven’t been listened. Therefore we design a recommendation system that can help coach to arrange appropriate music or songs also meet learners preference in music. The music that learners prefer can not only help learners to get into lesson more but also stimulate them to exercise and achieve a better effect by listening it. According to the result of the experiment of this recommendation system, the satisfaction of the learners has increased from 61.82% to 89.09%.This is a great proof of this system can achieve the goal of this research.