Comparison of content-based music recommendation using different distance estimation methods

• Published in: Applied intelligence (Dordrecht, Netherlands) Vol. 38; no. 2; pp. 160 - 174
• Main Author: Liu, Ning-Han
• Format: Journal Article
• Language: English
• Published: Boston Springer US 01.03.2013; Kluwer; Springer Nature B.V
• Subjects: Acoustics; Applied sciences; Approximation; Artificial Intelligence; Collaboration; Computer Science; Computer science; control theory; systems; Computer systems and distributed systems. User interface; Exact sciences and technology; Fundamental areas of phenomenology (including applications); Genetic algorithms; Genetics; Machines; Manufacturing; Mathematical analysis; Mechanical Engineering; Methods; Music; Musical performances; On-line systems; Perception; Physics; Processes; Programming; Ratings & rankings; Recommender systems; Similarity; Similarity measures; Software; Transduction; acoustical devices for the generation and reproduction of sound; Websites; Music recommendation; Maximum likelihood; Genetic algorithm; Distance function; Genetic programming; Complex function; Content management; Euclidean geometry; User centred design; Similarity; Customization; Case based reasoning; Similarity solution; Randomized algorithm; Experimental study; Recommendation; Preference; Music; Signal processing; Electronic trade; User behavior; Distance
• ISSN: 0924-669X; 1573-7497
• DOI: 10.1007/s10489-012-0363-y

With rapid growth in the online music market, music recommendation has become an active research area. In most current approaches, content-based recommendation methods play an important role. Estimation of similarity between music content is the key to these approaches. A distance formula is used to calculate the music distance measure, and music recommendations are provided based on this measure. However, people have their own unique tastes in music. This paper proposes a method to calculate a personalized distance measure between different pieces of music based on user preferences. These methods utilize a randomized algorithm, a genetic algorithm, and genetic programming. The first two methods are based on Euclidean distance calculation, where the weight of each music feature in the distance calculation approximates user perception. The third method is not limited to Euclidean distance calculation. It generates a more complex distance function to estimate a user’s music preferences. Experiments were conducted to compare the distance functions calculated by the three methods, and to compare and evaluate their performance in music recommendation.