Progress - 12/28

1. Trace file ：200 users, 4772 tracks,
each track has 4 attributes (aritst, album, track name, genre)

2. First experiment : content-based recommendation
a. Each user has 40 tracks and queries 10 tracks
b. Users exchange profiles when constructing neighborhood
c. When a user successfully queries one file from the other user, they exchange profiles
d. When a user receives other's queries, it recommends some tracks similar to the query file
e. Definition of similar tracks : more than two attributes are the same

3. Results:
There are totally 2000 queries.
If the users recommend based on content similarity, there are 381 tracks received
by users before they send queries. (373/2000 = 18.65%)

Progress - 12/17

1. Extract trace file from http://ws.audioscrobbler.com
2. Graph construction

Conception

1. Graph construction (based on peer similarity) +Profile exchange
2. Training (Query) + Construct relevance table between music
3. Recommendation
a. content-based : compare metadata similarity
b. collaborative-based : compare peer similarity
c. time-based : look up relevance table

Paper Survey

• Amazon.com Recommendations - item-to-item Collaborative Filtering (IEEE Computer Society 2003)
• Fab : content-based, collaborative recommendation (Communication of the ACM 1997)
  
• Content-based Collaborative Filtering for Improved Recommendations (American Association for Artificial Intelligence 2002)
• Distributed Collaborative Filtering for Peer-to-Peer File Sharing Systems (SAC'06)
• A user-oriented contents recommendation system in peer-to-peer architecture (Expert Systems with Application 2008)
• Distributed Collaborative Filtering with Domain Specialization (Resys'07)
• Personalized Recommendation Driven by Information Flow (SIGIR'06)
• Personalized of a Peer-to-Peer Television System