user_collab_code.txt

path <- "directory"
setwd(path)

# load library
library(data.table)
library(recommenderlab)
library(reshape2)
library(ggplot2)

# load and check data ---------------------------------------------------------------

train <- fread("train.csv")
test <- fread("test.csv")
sample_sub <- fread("sample_submission.csv")

head(train)
head(test)

str(train)
str(test)

# order data
train <- train[order(PID)]
test <- test[order(PID)]
summary(train)


# converting long format into wide format----------------------------------
train_dcast <- dcast(data = train, PID ~ Event,length, value.var = "Event")

#numeric conversion---------------------------------------------
a <-sapply(data.frame(train_dcast),as.numeric)
class(a)
is.na(a) <- !a
a[a== 0]<- NA

#user-collaborating-filtering-----------------------------
R<-as.matrix(a)
class(R)

r <- as(R, "realRatingMatrix")

r
as(r, "list")     # A list
as(r, "matrix")   # A sparse matrix

class(r)


rec_predicted <- Recommender(r[1:nrow(r)],method="UBCF")

print(rec_predicted)
model_details <- (getModel(rec_predicted))
getModel(rec_predicted)$data
model_details$data


recom_p <- predict(rec_predicted,r[1:10], type ="ratings", n=10)
recom_predict <- predict(rec_predicted,r[1:nrow(r)], n=10)
recom

list <- as(recom_p, "list")
as(r, "matrix")
a1 <- as(recom_p, "matrix")[,1:10]

# Convert all your recommendations to list structure

user_recom<-as.data.frame(list)

# get top 10 events per row
random_submit <- colnames(train_dcast)[-1][apply(train_dcast[,-c('PID'),with=F],1, function(x)order(-x)[1:10])]

# create the submission file
random_mat <- as.data.table((matrix(random_submit,ncol = 10, byrow = T)))
colnames(random_mat) <- colnames(sample_sub)[-1]
random_mat <- cbind(PID = test$PID, random_mat)
fwrite(random_mat,"random_sub.csv")

summary(train)