Smart Playlist Generation Using Data Mining Techniques

Abstract

Recommender systems are popular social web tools, as they address the information overload problem and provide personalization of results. This work presents a large-scale collaborative approach to the crucial part in the playlist generation process: next song recommendation. The purpose of this work is to evaluate the best suitable data mining functions, explore and express the best features of music. The focus will be on identifying the features to relate when do two songs sound alike? The human brain is for some reason designed to pick up on such similarities between individual tracks - or at least be trained to do so. The profusion of digital music files (i.e., mp3) on the internet and on personal hard disks makes it almost impossible for a music listener to know about and to sort out his/her music. Given the huge amount of music already available, users need support from technology to create their music playlists and to get more fun out of their music collection. We consider playlist creation as a shared responsibility between a user and function. The user indicates what kind of music should be included in a playlist by considering optimal attributes using a graphical user interface. The task of the development is then to find a 'best possible' playlist that satisfies all or most constraints. Smart playlist generation from a given set of constraints happens to be a problem that can be solved using data mining functionalities. In this the issues are with : Analysis and modeling of attributes chosen, Evaluation of modeled attributes for generation of playlist and also in identifying the similar category of songs, Generating a playlist of any length and music collections of any size and any variety. It should reflect the musical preference given by the user. Finally, our intent is to select optimal attributes, to identify and develop the best suitable Data Mining functionality that could be applied to this class of problem.