library(dplyr)
##
## Attaching package: 'dplyr'
## The following objects are masked from 'package:stats':
##
## filter, lag
## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, union
library(ggplot2)
set.seed(2018)
#make random data to cluster
synth.data <- data.frame(x1 = c(rnorm(20, 3, 1.5), rnorm(20,0,1), rnorm(20,5,1)),
x2 = c(rnorm(20, 0, 1), rnorm(20,4,1), rnorm(20,5,1)))
ndata <- nrow(synth.data)
ndim <- ncol(synth.data)
#plot the data
synth.data %>% ggplot(aes(x = x1, y= x2)) + geom_point(shape = 1) + theme_bw()

#Euclidean distance
u_dist <- function(u, v){
sqrt(sum((u - v) ** 2))
}
#we will make (k=3) k_means model.
k <- 3
cents <- data.frame(cl = 1:k)
# pick three coordinate
cents <- cbind(cents, synth.data[sample(1:60, k),])
cents
## cl x1 x2
## 27 1 -1.0586618 6.597463
## 30 2 -0.6318248 4.275945
## 33 3 -0.2662727 5.474087
#make it as a function, and iterate until it does not change.
# cent_f <- function(cents){
# #a: distance between cent1 and x,b: distance betweencent 2 and x, c: distance between cent3 and x.
# a <- apply(synth.data, 1, function(x){u_dist(cents[1,2:3],x)})
# b <- apply(synth.data, 1, function(x){u_dist(cents[2,2:3],x)})
# c <- apply(synth.data, 1, function(x){u_dist(cents[3,2:3],x)})
#
# #each row contains the information of distance between data and centroid.
# mat <- data.frame(dcent1 = a, dcent2 =b , dcent3 = c)
#
# mat_min <- apply(mat,1,which.min)
#
# #make a group using an index.
# gr1 <- synth.data [mat_min == 1,]
# gr2 <- synth.data [mat_min == 2,]
# gr3 <- synth.data [mat_min == 3,]
# # calculate mean in the gr 1,2,3 . and save it in the variable n_cents
# n_cents <- data.frame(cl = 1:3 , x1 = c(mean(gr1[,1]),mean(gr2[,1]),mean(gr3[,1])),x2 = c(mean(gr1[,2]),mean(gr2[,2]),mean(gr3[,2])))
# #compare all the elements in the matrix
# all(n_cents == cents)
#
#if centroid does not change, then stop.
# if ( all(n_cents == cents) ){
# synth.data$cl <- mat_min
# synth.data %>%
# ggplot(aes(x = x1, y= x2, col = cl)) + geom_point(shape = 1) + theme_bw() + geom_point(data = n_cents, shape = 4, col = 'red')
# }
# else{
# #if it changes, then do it again wiht new centroid. (recursive function.)
# cent_f(n_cents)
# }
# }
#
#
# mat_min
#
# cent_f(cents)
# mat_min
#this works at the R studio, but not when i kint this as html. so, just make another version with the same logic.
#to make mat min , we already know 4 times is optimal (from the previous one).
for (i in 1:4){
a <- apply(synth.data, 1, function(x){u_dist(cents[1,2:3],x)})
b <- apply(synth.data, 1, function(x){u_dist(cents[2,2:3],x)})
c <- apply(synth.data, 1, function(x){u_dist(cents[3,2:3],x)})
mat <- data.frame(dcent1 = a, dcent2 =b , dcent3 = c)
mat_min <- apply(mat,1,which.min)
gr1 <- synth.data [mat_min == 1,]
gr2 <- synth.data [mat_min == 2,]
gr3 <- synth.data [mat_min == 3,]
cents <- data.frame(cl = 1:3 , x1 = c(mean(gr1[,1]),mean(gr2[,1]),mean(gr3[,1])),x2 = c(mean(gr1[,2]),mean(gr2[,2]),mean(gr3[,2])))
}
mat_min
## [1] 2 2 2 2 2 2 3 2 2 2 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
## [36] 1 1 1 1 1 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3
synth.data$cl <- mat_min
synth.data %>%
ggplot(aes(x = x1, y= x2, col = cl)) + geom_point(shape = 1) + theme_bw() + geom_point(data = cents, shape = 4, col = 'red' )

library(recommenderlab)
## Loading required package: Matrix
## Loading required package: arules
##
## Attaching package: 'arules'
## The following object is masked from 'package:dplyr':
##
## recode
## The following objects are masked from 'package:base':
##
## abbreviate, write
## Loading required package: proxy
##
## Attaching package: 'proxy'
## The following object is masked from 'package:Matrix':
##
## as.matrix
## The following objects are masked from 'package:stats':
##
## as.dist, dist
## The following object is masked from 'package:base':
##
## as.matrix
## Loading required package: registry
library(dplyr)
library(tidyr)
##
## Attaching package: 'tidyr'
## The following object is masked from 'package:Matrix':
##
## expand
library(tibble)
library(doBy)
load('ratings.RData')
#Pearson correlationfunction.
pearsonCor <- function(x, y){
x_mean <- mean(x, na.rm = T)
y_mean <- mean(y, na.rm = T)
idx <- !is.na(x) & !is.na(y)
if(sum(idx) == 0) return(NA)
x_new <- x[idx]
y_new <- y[idx]
sum((x_new- x_mean) * (y_new -y_mean)) / sqrt( sum( (x_new - x_mean)**2) * sum( (y_new-y_mean) **2) )
}
recommend <- function(u){
sim <- apply(ratings, 1, function(x) {
pearsonCor(u, x)
})
sim
k = 2
k_neighbors <- setdiff(which.maxn(sim, k+1), 5)
k_recommend <- apply(ratings[k_neighbors,], 2, function(x) { mean(x, na.rm = T)})
k_recommend
k_recommend_final <- k_recommend[is.na(u)]
print(sort(k_recommend_final, decreasing = T))
}
#Apply it to all users
apply(ratings,1, function(x){recommend(x)})
## Me Before You Sherlock
## 4 3
## Sherlock Titanic Matrix
## 5 2 1
## The Avengers Me Before You
## 5 1
## Titanic The Avengers Transformer
## 5 3 2
## The Avengers Titanic
## 5 2
## Titanic Transformer Me Before You
## 5 3 1
## $A
## Me Before You Sherlock
## 4 3
##
## $B
## Sherlock Titanic Matrix
## 5 2 1
##
## $C
## The Avengers Me Before You
## 5 1
##
## $D
## Titanic The Avengers Transformer
## 5 3 2
##
## $E
## The Avengers Titanic
## 5 2
##
## $F
## Titanic Transformer Me Before You
## 5 3 1
ratings.df <- read.csv('train_v2.csv')
#delete user id, and move user to rowname.
rating_matrix <- ratings.df %>%
select(-ID) %>%
spread(movie, rating) %>%
remove_rownames() %>%
column_to_rownames(var = 'user')
#make it as continuous data type
rating_rrm <- as(as(rating_matrix, 'matrix'), 'realRatingMatrix')
# row name : movie name, col name : users
# I picked the movies which have more than 51 reviews,and users which rate over 100 reviews.
rating_rrm <- rating_rrm[rowCounts(rating_rrm) > 50, colCounts(rating_rrm) > 100]
#normalize the data(becuase there would be some poeple who would like to give it only biased score, 4,5 . or 1,2)
rating_rrm_norm <- normalize(rating_rrm)
image(rating_rrm[1:50, 1:50])

image(rating_rrm_norm[1:50, 1:50])

print(object.size(rating_matrix), units = "auto")
## 85.1 Mb
print(object.size(rating_rrm), units = "auto")
## 7.5 Mb
recommenderRegistry$get_entries(dataType = "realRatingMatrix")
## $ALS_realRatingMatrix
## Recommender method: ALS for realRatingMatrix
## Description: Recommender for explicit ratings based on latent factors, calculated by alternating least squares algorithm.
## Reference: Yunhong Zhou, Dennis Wilkinson, Robert Schreiber, Rong Pan (2008). Large-Scale Parallel Collaborative Filtering for the Netflix Prize, 4th Int'l Conf. Algorithmic Aspects in Information and Management, LNCS 5034.
## Parameters:
## normalize lambda n_factors n_iterations min_item_nr seed
## 1 NULL 0.1 10 10 1 NULL
##
## $ALS_implicit_realRatingMatrix
## Recommender method: ALS_implicit for realRatingMatrix
## Description: Recommender for implicit data based on latent factors, calculated by alternating least squares algorithm.
## Reference: Yifan Hu, Yehuda Koren, Chris Volinsky (2008). Collaborative Filtering for Implicit Feedback Datasets, ICDM '08 Proceedings of the 2008 Eighth IEEE International Conference on Data Mining, pages 263-272.
## Parameters:
## lambda alpha n_factors n_iterations min_item_nr seed
## 1 0.1 10 10 10 1 NULL
##
## $IBCF_realRatingMatrix
## Recommender method: IBCF for realRatingMatrix
## Description: Recommender based on item-based collaborative filtering.
## Reference: NA
## Parameters:
## k method normalize normalize_sim_matrix alpha na_as_zero
## 1 30 "Cosine" "center" FALSE 0.5 FALSE
##
## $POPULAR_realRatingMatrix
## Recommender method: POPULAR for realRatingMatrix
## Description: Recommender based on item popularity.
## Reference: NA
## Parameters:
## normalize
## 1 "center"
## aggregationRatings
## 1 new("standardGeneric", .Data = function (x, na.rm = FALSE, dims = 1,
## aggregationPopularity
## 1 new("standardGeneric", .Data = function (x, na.rm = FALSE, dims = 1,
##
## $RANDOM_realRatingMatrix
## Recommender method: RANDOM for realRatingMatrix
## Description: Produce random recommendations (real ratings).
## Reference: NA
## Parameters: None
##
## $RERECOMMEND_realRatingMatrix
## Recommender method: RERECOMMEND for realRatingMatrix
## Description: Re-recommends highly rated items (real ratings).
## Reference: NA
## Parameters:
## randomize minRating
## 1 1 NA
##
## $SVD_realRatingMatrix
## Recommender method: SVD for realRatingMatrix
## Description: Recommender based on SVD approximation with column-mean imputation.
## Reference: NA
## Parameters:
## k maxiter normalize
## 1 10 100 "center"
##
## $SVDF_realRatingMatrix
## Recommender method: SVDF for realRatingMatrix
## Description: Recommender based on Funk SVD with gradient descend.
## Reference: NA
## Parameters:
## k gamma lambda min_epochs max_epochs min_improvement normalize verbose
## 1 10 0.015 0.001 50 200 1e-06 "center" FALSE
##
## $UBCF_realRatingMatrix
## Recommender method: UBCF for realRatingMatrix
## Description: Recommender based on user-based collaborative filtering.
## Reference: NA
## Parameters:
## method nn sample normalize
## 1 "cosine" 25 FALSE "center"
#User based CF
ubcf_model <- Recommender(rating_rrm, method = 'UBCF')
#모델과 실제를 본다.
recom <- predict(ubcf_model, rating_rrm[1:2, ])
recom <- predict(ubcf_model, rating_rrm[1:2, ], type = 'ratings')
recom
## 2 x 1790 rating matrix of class 'realRatingMatrix' with 3364 ratings.
# divide into two sets, train and test
e <- evaluationScheme(rating_rrm, method="split", train=0.8, given=10)
#r1: user based. r2: item based.
r1 <- Recommender(getData(e, "train"), "UBCF")
r2 <- Recommender(getData(e, "train"), "IBCF")
#which one is more accurate?
p1 <- predict(r1, getData(e, "known"), type="ratings")
p2 <- predict(r2, getData(e, "known"), type="ratings")
error <- rbind( calcPredictionAccuracy(p1, getData(e, "unknown")),
calcPredictionAccuracy(p2, getData(e, "unknown"))
)
rownames(error) <- c("UBCF","IBCF")
error
## RMSE MSE MAE
## UBCF 1.064338 1.132815 0.8447415
## IBCF 1.758544 3.092475 1.3555584