Book excerpt: "The development of the Internet and the emergence of audio compression technologies have contributed to the realization of making millions of music titles accessible to millions of users. Due to the extensive distribution of music, consumers are being presented with a problem of information overload, while the music industry is being faced with the challenge of personalized promotion and distribution. Music recommendation systems aim to ease the task of finding the music items that might interest the users by generating meaningful recommendations. The recommendation for music is different from those for books and movies, due to its low cost per item, short consumption time, high per-item reuse, highly contextual usage, and numerous item types. Understanding the patterns of music listening and consumption is important to create accurate and satisfying music recommendations. This thesis reviews state-of-the-art music recommendation and discovery methods with the goal of presenting the historical developments in this area. Traditional music recommendation systems can be classied as one of two major kinds: collaborative filtering and content-based filtering. Recently, the research community has broadened its attention to include other aspects, such as hybrid approaches, context awareness, social tagging, music networks, visualization, playlist generation, and group recommendation. For the evaluation of music recommendation systems, researchers or developers need to take into account properties such as accuracy, coverage, confidence, novelty, diversity, and privacy. These properties can be measured in an offline simulation, a user study, or an online evaluation. Suggestions for future work in both the design and the evaluation of music recommendation systems are given." --