Optimize CSV reader

This code takes more than 5 days running, but, it worked perfectly on the data test.

T1=Sys.time()
# Recommendation Program
#plus proches voisins
#install.packages("reshape2") like require(reshape2)
#require(data.table)
#require(reshape2)
#install.packages("data.table")
#library(Rserve)
library(data.table)
library(reshape2)


table1 <- read.table("base_test_fil.csv", sep=";", quote="\"",  header=TRUE)
#fread(data.table)

table3=as.data.table(table1)
#aggregation + binary matrix of table1
table2 <- dcast.data.table(table3, id_customer ~ product_id,value.var="product_id", fun.aggregate=length)
table2=as.data.frame(table2)
#verification
#class(table2)
data.stat=table2

#remark: we can do directly :data.stat=table2 
options(digits=4)

# Drop any column named "id_order":technique :not in 
data.stat.ibs <- (data.stat[,!(names(data.stat) %in% c("id_customer"))])

# Create a placeholder dataframe listing item vs. item
data.stat.ibs.similarity  <- matrix(NA, nrow=ncol(data.stat.ibs),ncol=ncol(data.stat.ibs),dimnames=list(colnames(data.stat.ibs),colnames(data.stat.ibs)))

getCosine <- function(x){
   as.matrix(t(x[,c(1:ncol(x))])%*%x[,c(1:ncol(x))]/(sqrt(colSums(x^2) %*% t(colSums(x^2)))))
}

data.stat.ibs.similarity<-getCosine(as.matrix(data.stat.ibs))

# Back to dataframe
data.stat.ibs.similarity <- as.data.frame(data.stat.ibs.similarity)
#View(data.stat.ibs.similarity)

# Get the top 10 neighbours for each
data.stat.neighbours <- matrix(NA, nrow=ncol(data.stat.ibs.similarity),ncol =11,dimnames=list(colnames(data.stat.ibs.similarity)))

for(i in 1:ncol(data.stat.ibs)) 
{
    data.stat.neighbours[i,] <- (t(head(n=11,rownames(data.stat.ibs.similarity[order(data.stat.ibs.similarity[,i],decreasing=TRUE),][i]))))
}

data=data.stat.neighbours
data=data[,-1]
colnames(data)=c("best1","best2","best3","best4","best5","best6","best7","best8","best9","best10")

#Export csv file 
write.csv(data,file="export1.csv",row.names=TRUE)
#Sys.time()-T1
#####user_based_collaborative
#Lets make a helper function to calculate the scores
getScore <- function(history, similarities)
{
  x <- sum(history*similarities)/sum(similarities)
  x
}

# A placeholder matrix
holder <- matrix(NA,nrow=nrow(data.stat),ncol=ncol(data.stat)-1,dimnames=list((data.stat$id_customer),colnames(data.stat[-1])))

# Loop through the id_customers (rows)
for(i in 1:nrow(holder)) 
{
  # Loops through the products (columns)
  for(j in 1:ncol(holder)) 
  {
    # Get the id_customer's name and th product's name
    # We do this not to conform with vectors sorted differently 
    id_customer <- rownames(holder)[i]
     product <- colnames(holder)[j]

     # We do not want to recommend products you have already consumed
     #Nous ne voulons pas de recommander des produits que vous avez déjà consommés
     # If you have already consumed it, we store an empty string
     if(as.integer(data.stat[data.stat$id_customer==id_customer,product]) == 1)
     { 
       holder[i,j]<-""
     } else {
       # We first have to get a product's top 10 neighbours sorted by similarity
       topN<-((head(n=11,(data.stat.ibs.similarity[order(data.stat.ibs.similarity[,product],decreasing=TRUE),][product]))))
       topN.names <- as.character(rownames(topN))
       topN.similarities <- as.numeric(topN[,1])

       # Drop the first one because it will always be the same product
       topN.similarities<-topN.similarities[-1]
       topN.names<-topN.names[-1]

       # We then get the id_customer's purchase history for those 10 items
       topN.purchases<- data.stat[,c("id_customer",topN.names)]
       topN.id_customerPurchases<-topN.purchases[topN.purchases$id_customer==id_customer,]
       topN.id_customerPurchases <- as.numeric(topN.id_customerPurchases[!(names(topN.id_customerPurchases) %in% c("id_customer"))])

       # We then calculate the score for that product and that id_customer
       holder[i,j]<-getScore(similarities=topN.similarities,history=topN.id_customerPurchases)

    } # close else statement
  } # end product for loop   
} # end id_customer for loop

# Output the results to a file
data.stat.id_customer.scores <- holder
#write.csv(file="final-id_customer-scores.csv",data.stat.id_customer.scores)

# Lets make our recommendations pretty
data.stat.id_customer.scores.holder <- matrix(NA,   nrow=nrow(data.stat.id_customer.scores),ncol=100,dimnames=list(rownames(data.stat.id_customer.scores)))
for(i in 1:nrow(data.stat.id_customer.scores)) 
{
  data.stat.id_customer.scores.holder[i,] <- names(head(n=100,(data.stat.id_customer.scores[,order(data.stat.id_customer.scores[i,],decreasing=TRUE)])[i,]))
}

# Write output to file
write.csv(data.stat.id_customer.scores.holder,file="export3.csv",row.names=TRUE)