CF - SlopeOne Implementation in R

gistfile1.r

#########################################################################
#-----------------------------------------------------------------------#
#         SlopeOne Recommender Implementation in R                      #
#########################################################################

#####################
#    Sample Data
#####################

# sample data 1
Mov1 <- c(4,4,3,4,2)
Mov2 <- c(NA,2,NA,4,1)
Mov3 <- c(5,1,2,NA,3)
Mov4 <- c(5,NA,4,NA,5)

username = paste("U",1:5,sep="")
t <- paste("Item",1:7, sep="")

item <- data.frame(Mov1,Mov2,Mov3,Mov4, row.names=username)
user <- t(item)

# sample data 2 (Mahout in Action)
U1 <- c(5.0, 3.0, 2.5, NA, NA, NA, NA)
U2 <- c(2.0, 2.5, 5.0, 2.0, NA, NA, NA)
U3 <- c(2.5, NA, NA, 4.0, 4.5, NA, 5.0)
U4 <- c(5.0, NA, 3.0, 4.5, NA, 4.0, NA)
U5 <- c(4.0, 3.0, 2.0, 4.0, 3.5, 4.0, NA)

data <- data.frame(U1,U2,U3,U4,U5, row.names=paste("Item",1:7,sep=""))



#####################################
# SlopeOne Difference Calculation
#####################################

Itemdata <- t(data)
Userdata <- data

diff_xy <- function(x,y) {
  n_corate <- length(which(x&y))
  if(n_corate == 0){
    avg_diff <- NA
  }
  else {
    avg_diff <- sum(x-y,na.rm=TRUE)/n_corate
  }
  return(avg_diff)
}


slopeone_diff <- function(x) {
  m <- matrix(NA, nrow=ncol(x),ncol=ncol(x),dimnames=list(colnames(x),colnames(x)))
  diff <- as.data.frame(m)
  for(i in 1:ncol(x)) {
    for(j in i:ncol(x)) {
      if(i==j){
        diff[i,j]=NA
        diff[j,i]=NA
      }
      else {
        diff[i,j]= diff_xy(x[,i],x[,j])
        diff[j,i]= -1 * diff[i,j]
      }
    }
  }
  return(diff)
}

slopeonediff <- slopeone_diff(Itemdata)
                   
                   
#####################################
#         Recommendation
#####################################

slopeone_reco <- function(userdata,diffdata) {
  m <- matrix(NA, nrow=nrow(userdata),ncol=ncol(userdata),dimnames=list(colnames(t(userdata)),colnames(userdata)))
  reco <- as.data.frame(m)
  # loop over users
  for(i in 1:ncol(userdata)) {
    # vector of a user's rated items and their corresponding vector index
    rated <- c(na.omit(userdata[,i]))
    ind_rated <- which(!is.na(userdata[,i]))
    ind_unrated <- which(is.na(userdata[,i]))  
    n_unrated <- length(userdata[is.na(userdata[,i])])
    # loop over nonrated items of a user, if there is at least one unrated item
    if(n_unrated > 0){
      for(j in 1:length(ind_unrated)){      
        reco[ind_unrated[j],i] = sum(rated - diffdata[ind_rated,ind_unrated[j]],na.rm=TRUE) / length(rated)
      }
    }  
  }
  return(reco)
}

my.slopeone.reco <- slopeone_reco(Userdata,slopeonediff)