a Recommender system Based on Genetic Algorithm for Music Data Hyun-Tae Kim, Eungyeong Kim, Jong-Hyun Lee, Chang Wook Ahn School of Information Communication Engineering Sungkyunkwan University (SKKU) Cheoncheon-dong, Suwon 440-746, SKorea *cwan @skku. edu(corresponding author Abstract -Nowadays, recommender systems are widely In this paper, a new recommender system for implemented in E-commerce websites to assist customers in music data b g the content-based filter finding the items they need A recommender system should technique with active genetic algorithm. We also be able to provide users with useful information about the consider the unique properties of each music track, such as items that might interest them. The ability of promptly tempo, pitch and chord. We use a music feature extraction responding to changes in user's preference is a valuable as tool to analyze these properties. The results of the extraction for such systems. This paper presents an innovative consist in the database of our proposed system. We expect recommender system for music data that combines two that the proposed system will provide more suitable methodologies, the content-based filtering technique and the information, which adapts to the preference of each user, by interactive genetic algorithm. The proposed system aims to effectively adapt and respond to immediate changes in use applying the genetic algorithm to the system. The user ferences. The experiments conducted in an objective pref manner exhibit that our system is able to recommend items acquiring records, the recommender system analyzes and suitable with the subjective favorite of each individual user. recommends items that are appropriate with their own favorite mender system; user's prej This paper is organized as follow. Section II reviews interactive genetic algorithm; content-based related work. Section ii describes the structure of our recommender system and explains how to op erate genetic algorithm in this system. In Section IV, we provide In daily life it is often necessary to make a decision experimental results and analysis. Finally, Section V without resort to enough personal experience of various concludes this paper alternatives. When the alternatives domain is quite large, it is difficult for users to make an appropriate decision. For IL. RELATED WORK example, many people might rely on recommendations from A. Recommender Systems knowledge or advertisements, and reviews about and book in magazines. Such diverse references ma The main issue of a recommender system is how to recommend items tailored with users preference from Ip the users in making a proper choice. In this regard, a resources. The recommender system also has to recognize recommender system has the same usefulness, but provides the users with a refined list of alternatives tailored with their and provide items corresponding with favorite of users. In own preferences [1],[2 order to resolve this matter, there are two main approaches Recommender systems are useful for people living in collaborative filtering and content-based filtering [4] these days. After the 1990s of the 20th century, the Internet In the collaborative filtering approach, the recommender technologies, especially World Wide Web, have grown with system provides recommendation by collecting users astonishing speed. With this change, there are numerous profiles and discovers relations between each profile. After resources, such as document, photo and music data, which identifying correlation of each profile, the system classifies are accessible on the Internet. Many of end users having profiles that are similar to the others. The customers will face the problem that which resource is more ystem then recommends items derived from other profiles suitable than the others. Many of the largest commerce in the same group. The advantage of this approach, thus, is websites, thus, have already been using recommender that it has a high possibility to recommend items tems to assist their customers in searching items they corresponding with users preference by providing ould like to purchase, such as the b-commerce web site environments in which each user can share his or her own Amazon.comandthesearchengineofGooglecomTheseprofile[51,[6] In the altermative approach, the content-based filt systems provide with the search results tailored to users own the recommender system examines the description of the preference 3] items which are rated higher than others from users. After 978-1-4244-6349-7/10/$26.00@2010EEE V6414           !   "  #$ %& ' (   )$   !     & &. , -* ,+ $    . 00/201 3 '. 4&&3*5  +                                                                              !                                             "                                       !        #        $     #                 $                                      (3 (6 7,$ (76 (        &   .     5  85   -  -3%  -  9     &  555  3 : 8 5  55      ; & .  -     -. < - <& = 3 -    5 & 553(           < 5- .  -. . 5 @>A@?A3         55 -   3>BB/?/ (    5%%%<-. .    53 %              5     .  <   (  3         .5< .  <   3        .<  -  <                   . &  5    !  .<   = 3         3 3   5-.;. 5 @CA3 ( 55.55 .      <  <      <    9 .   -    3 %    9 55   &    5 5   3%     8  =553 8    <   55  3% 85   55  . 5-  <    .55 < 55         3   5   <     3  9         =          555 .  . -3  55   =  .3  (( -.  .&3  ((( <           85  .  5        3 (  (D.5- 85       3 :   D  553 ((3 !# !%7                  .        . ; 5   3          =   5-   5   . -  3 (  - .  55E          @0A3 ( <-  55    5-    <   ; 5   -   <.  53        5    -  5       3        -   5     53  - 55      5<       5   . ; 5  < 5-   -   .        . 5@FA@1A3 (   - 55    <        8    5     .       3  978-1-4244-6349-7/10/$26.00 c 2010 IEEE V6-414