This paper presented a novel approach in recommending songs one by one
based on user behavior. The approach considered genre, recording year,
freshness, favor and time pattern as factors to recommend songs. The evaluation
results demonstrate that the approach is effective.