RecommendEngine

Implement the Item-based Recommendation Algorithm and calculate the similarity of users and items.

Notice : I'm sorry about that it is not efficient enough in this recommendation engine when calculating the similarity matrix and computing the prediction matrix. I have optimized this problem and the code will be shared soon.

Implementions

• Item-based Recommendation Algorithm is the most commomly used algorithm in Recommendation System, you can refer to 《Mahout in Action》Chapter1.6 “Distributing recommendation computations” for more details.
• I calculated the Item-Similarity and User-Similarity by reading paper “Empirical Analysis of Predictive Algorithms for Collaborative Filtering”， by John S.Breese.
• I also implement a CRON for RecommendEngine so that the recently-records can be fetched from MySQL to HDFS and after accomplish the computation tasks, the recommended result will be loaded into MySQL periodically.

DataSets

• I got MovieLens DataSets from here.

Running Environment

• Ubuntu 14.04
• JDK 1.7.0_75
• Hadoop 1.2.1
• Sqoop 1.3.0

Feedback

• Email: [email protected] or [email protected]
• Sina Weibo: @tinylcy