Amazon like Ecommerce site and Recommendation system

4
votes

I am planning to implement a basic recommendation system that uses Facebook Connect or similar social networking site API's to connect a user's profile, do an analysis based on tags and use the results to generate item recommendations on my e-commerce site (works similar to Amazon).

I do believe I need to divide parts into such:

  1. Fetching social networking data via API's.(Indeed user allows this)

  2. Analyze these data and generate tokes.

  3. By using information tokens, do item recommendations on my e-commerce site.

EG: I am a fan of "The Strokes" band on my Facebook account, system analyzes this and recommends me "The Strokes Live" CD.

For any part(fetching data, doing recommendation based on tags...), what algorithm and method would you recommend/ is used?

2
social-networkingrecommendation-engine
Is your ecommerce cite already live? It helps a lot if you already have data based on your users. Essentially you will want to train a model based on the idea that uses who bought x have these features(tags). If no one has bought anything yet, its tough... - frankc

2 Answers

8
votes

Good practical books on these kind of algorithms are:

2
votes

Being on your place I would definitely had a look at the algorithms and articles published during the Netflix contest. See forum and the sites of the best teams.