適性化飛輪音樂推薦系統之設計與研究-基於使用者喜好

新媒體與飛輪運動之興起，大幅提升購買飛輪車回家的意願，其中，音樂喜好會影響騎乘動力，然而多數飛輪課程使用之音樂皆由教練安排，無法依照使用者喜好挑選，因此本研究欲探討飛輪音樂推薦系統是否能讓使用者在家更有動力騎乘飛輪。本研究之飛輪音樂推薦系統將使用者課後的偏好評分，與事先蒐集之音樂特徵值結合，並使用單純貝氏分類器，計算出使用者對歌曲的喜愛程度，而後透過輪盤法推薦音樂，並結合教練安排之動作，進而產生使用者音樂喜好之飛輪課程；將受測者分為系統組及對照組以此來觀察使用系統所帶來的影響，經由實驗結果發現，喜好音樂推薦之準確度不理想。此外結合計畫行為理論與科技接受模式，分析使用者對於適性化飛輪音樂推薦系統之接受程度，經由分析問卷得出，多數使用者能投入於課程中，並提升運動動力。根據結果分析顯示，整體接受聲量比預期高，若修改此系統，將準確度提高，必定系統接受度更高，讓使用者更加有動力，不會因地自限更有運動的意願，並達到運動效果。

關鍵字

偏好評分；單純貝氏分類器；音樂推薦機制；飛輪運動；計畫行為理論；科技接受模式

並列摘要

The purchase intention of flywheel has greatly increased by the rise of new media and indoor spin workout. During the workout, song preferences will affect workout motivation. However, the music used in most spin workout courses are selected by the instructor and cannot be selected by the users. Therefore, this study attempts to explore whether the music recommendation system for spin workout can strengthen users’ workout motivation at home. The flywheel music recommendation system of this study combines the user’s preference score after class with the music eigenvalues collected in advance. Use a Naive Bayes Classifier to calculate the user’s preference for the music. Then recommend music through the roulette method, combined with the actions arranged by the instructor, then generate a flywheel course of the user's music preferences. The subjects were divided into a system group and a control group to observe the impact of using the system. According to the experimental results, it was found that the accuracy of favorite music recommendations was not ideal. In addition, analyzes the user's acceptance of the adaptive flywheel music recommendation system based on the Theory of Planned Behavior and Technology Acceptance Model. According to the analysis of the questionnaire, most users can participate in the course and improve their motivation. According to the analysis of the results, the overall received acceptance is higher than expected. If this system is modified, the accuracy will be improved, and the acceptance of the system must be higher. Let the user be more motivated, will not be more willing to exercise due to local constraints, and achieve the exercise effect.