张丹在这篇博客http://blog.fens.me/r-mahout-usercf/里给出了协同过滤算法的R代码，但是其本身使用过多的for循环，影响了代码的执行效率，同时也没有把R语言的特色给体现出来。数据在他的博客中下载。
<br />
# 1). 建立数据模型<br />
library(reshape2)<br />
file = 'c:/temp/testCF.csv'</p>
<p>FileDataModel = function(file){<br />
                  data = read.csv(file,header=FALSE)<br />
                  names(data) = c("uid","iid","pref")<br />
                  M = dcast(uid ~ iid,data=data)<br />
                  M = as.matrix(M[-1])<br />
                  rownames(M) = unique(data$uid)<br />
                  colnames(M) = unique(sort(data$iid))<br />
                  M<br />
                  }</p>
<p>M = FileDataModel(file)<br />
  101 102 103 104 105 106 107<br />
1 5.0 3.0 2.5  NA  NA  NA  NA<br />
2 2.0 2.5 5.0 2.0  NA  NA  NA<br />
3 2.5  NA  NA 4.0 4.5  NA   5<br />
4 5.0  NA 3.0 4.5  NA   4  NA<br />
5 4.0 3.0 2.0 4.0 3.5   4  NA<br />
#原文是把NA替换为0，但是考虑到后面计算的方便，这里把它保留了。<br />
# 2). 欧氏距离相似度算法<br />
 EuclideanDistanceSimilarity = function(M){<br />
                          distance = function(x,y){<br />
                                 nb = sum(!is.na(M[x,] - M[y,]))<br />
                                 s = sum((M[x,] - M[y,])^2,na.rm = T)<br />
                                 s = nb/(1 + sqrt(s))<br />
                                 if(s > 1) s = 1<br />
                                 if(s < -1) s = -1<br />
                                 return(s)<br />
                                 }<br />
                         outer(1:nrow(M),1:nrow(M),Vectorize(distance)) - diag(nrow(M))<br />
                         }<br />
S = EuclideanDistanceSimilarity(M)<br />
S<br />
          [,1]      [,2]      [,3]      [,4]      [,5]<br />
[1,] 0.0000000 0.6076560 0.2857143 1.0000000 1.0000000<br />
[2,] 0.6076560 0.0000000 0.6532633 0.5568464 0.7761999<br />
[3,] 0.2857143 0.6532633 0.0000000 0.5634581 1.0000000<br />
[4,] 1.0000000 0.5568464 0.5634581 0.0000000 1.0000000<br />
[5,] 1.0000000 0.7761999 1.0000000 1.0000000 0.0000000<br />
# 3). 最紧邻算法<br />
NearestNUserNeighborhood = function(S,n){<br />
                                     sapply(1:n,function(i) {m = apply(S,2,which.max)<br />
                                                        ij = cbind(m,seq_along(m))<br />
                                                        S[ij] <<- 0 # 这里要感谢颜站长，否则不知道<<-怎么用<br />
                                                        m})<br />
                                    }<br />
NEIGHBORHOOD_NUM = 2<br />
N = NearestNUserNeighborhood(S,NEIGHBORHOOD_NUM)<br />
N<br />
     [,1] [,2]<br />
[1,]    4    5<br />
[2,]    5    3<br />
[3,]    5    2<br />
[4,]    1    5<br />
[5,]    1    3<br />
# 4). 推荐算法，需要改进的是原文给出的输出是矩阵，这里的输出是向量。<br />
RECOMMENDER_NUM = 3<br />
UserBasedRecommender = function(uid,n,M,S,N){<br />
                        unuse = colnames(M)[is.na(M[uid,])]<br />
                        m = S[uid, N[uid,]]<br />
                        md = M[N[uid,],unuse]<br />
                        if(length(dim(md)) >= 1) {<br />
                                        a = colnames(md)[colSums(is.na(md)) == 0]<br />
                                        score = colSums(md*m,na.rm = T)/apply(!is.na(md),2,function(x) sum(m[x]))<br />
                                        res = score[order(score,decreasing=T)][1:n]<br />
                                        res = res[intersect(names(res),a)]<br />
                                        return(res)<br />
                                       } else { res = NA}<br />
                       }<br />
# 5). 运行程序</p>
<p>R1 = UserBasedRecommender(1,RECOMMENDER_NUM,M,S,N);<br />
R1<br />
104  106<br />
4.25 4.00</p>
<p>R2 = UserBasedRecommender(2,RECOMMENDER_NUM,M,S,N);<br />
R2<br />
    105<br />
3.956999</p>
<p>R3 = UserBasedRecommender(3,RECOMMENDER_NUM,M,S,N);<br />
R3<br />
     103      102<br />
3.185407 2.802432</p>
<p>R4 = UserBasedRecommender(4,RECOMMENDER_NUM,M,S,N);<br />
R4<br />
102<br />
  3 </p>
<p>R5 = UserBasedRecommender(5,RECOMMENDER_NUM,M,S,N);<br />
R5<br />
[1] NA<br />
</p>