A mood‐ and situation‐based model for developing intuitive Pop music recommendation systems

Traditionally, a per‐song‐purchased base recommendation system is used on most music websites, but this method produces unsatisfactory results under various situational practices. This study proposes a hybrid procedure that includes both an expert‐attributes selection capability and a mood/situation‐attributes categorization functionality. This procedure fosters the development of a so‐called MoMusic model as an unlimited online streaming service to replace current systems and artfully provide music to interested parties. This study employs a dataset consisting of 821 songs from the 2005–2010 annual music rankings as well as songs from the top artists from 2009 to 2010 from Taiwan's popular KKBOX music streaming service. The experimental dataset was assessed and coded by domain experts, and the expert‐attributes selections and mood/situation‐attributes categorizations were used to produce recommendation lists. These recommendation lists were then paired with questionnaire‐derived music preferences from experienced users. The experimental results conclusively show that the MoMusic model is approximately twice as accurate as the random selection‐based lists and the KKBOX‐like recommendation lists and performs better than the two listed recommendation systems. The MoMusic model scores 0.889 on the usefulness evaluation, whereas the system satisfaction is 0.96. The MoMusic model has the advantages of intuitive use and high performance.