Without the denominator, the prediction is not a weighted average -- it's some kind of weighted sum. The values will not be in nearly the same range as the input ratings -- might be in the thousands. It's not a prediction anymore. You can rank on it, but it will just favor items that co-occur with a lot of other items. The user ratings have almost no effect. It's just an unpersonalized result returning most popular items then. I don't know what Hulu does but I don't imagine this makes sense as a basis for a recommender.
On Wed, Jul 18, 2012 at 12:15 PM, Zhang YaJun(产品技术) <[email protected] > wrote: > Hi, > > There is a question about cf in Mahout. It’s about the formula prediction, > in your source it’s Prediction(u,i) = sum(all n from N: similarity(i,n) * > rating(u,n)) / sum(all n from N: abs(similarity(i,n))), this is in class > org.apache.mahout.cf.taste.hadoop.item.AggregateAndRecommendReducer. > > However, some companies as Hulu ,this formula is Prediction(u,i) = sum(all > n from N: similarity(i,n) * rating(u,n)), > > Why do you give the denominators, is there any problem? > > Thank you! > > Yajun Zhang > > www.sohu.com<http://www.sohu.com> > > > > 张亚军 > SOHU视频 > 产品技术-DM数据组 > 电话：010-61134332 > 手机：13699151315 > 地址：北京市海淀区王庄路一号清华同方科技广场D座西楼7层 > >
