, Thiago Lima [ctb],
Lucas Oliveira [ctb]| Title: | Recommendation by Collaborative Filtering |
|---|---|
| Description: | Provides methods and functions to implement a Recommendation System based on Collaborative Filtering Methodology. See Aggarwal (2016) <doi:10.1007/978-3-319-29659-3> for an overview. |
| Authors: | Jessica Kubrusly [aut, cre]
, Thiago Lima [ctb],
Lucas Oliveira [ctb] |
| Maintainer: | Jessica Kubrusly <[email protected]> |
| License: | GPL-3 |
| Version: | 0.3.0 |
| Built: | 2024-11-05 04:57:05 UTC |
| Source: | https://github.com/cran/CFilt |
CF is a class of objects that stores information about a recommendation system. This information includes the consumption or rating of each (user, item) pair in the utility matrix MU, the similarities between each pair of users in the similarity matrix SU, the similarities between each pair of items in the similarity matrix SI, the number of items consumed and/or rated by each user in the vector n_aval_u, the number of users who consumed and/or rated each item in the vector n_aval_i, the average rating value of each user in the vector averages_u, the average rating value received by each item in the vector averages_i, the number of items consumed in common by each pair of users in the matrix Int_U, and the number of users in common for each pair of items in the matrix Int_I. The class contains methods such as addNewUser, addNewEmptyUser, deleteUser, addNewItem, addNewEmptyItem, deleteItem, newRating and deleteRating, which modify the object's structure by altering users, items, or consumption data. The class also includes functions such as kClosestItems, topKUsers, and topKItems, which return items to recommend to a user or users to whom an item should be recommended. An object of the CF class is created using the CFBuilder function.
MUThe Utility Matrix, a matrix that contains all the users' ratings. The rows comprise users and the columns, items.
SUThe user similarity matrix.
SIThe item similarity matrix
IntUA symmetric matrix that records the number of users in common who consumed each pair of items.
IntIA symmetric matrix that records the number of items in common that have been consumed by each pair of users.
averages_uA vector that contains the averages of users' ratings.
averages_iA vector that contains the averages of items' ratings.
n_aval_uA vector that stores the number of items rated by each user.
n_aval_iA vector that stores the number of users who consumed each item.
datatypeA character that indicates the type of data, which can be either "consumption" or "rating".
addnewemptyitem(Id_i)A method that adds a new item that has not yet been consumed by any existing user in the recommendation system. Id_i: a character, the new item ID; To add more than one new user, lists can be used. Id_i: a list of characters;
addnewemptyuser(Id_u)A method that adds a new user who has not yet consumed any existing items in the recommendation system. Id_u: a character, the new user ID; To add more than one new user, lists can be used. Id_u: a list of characters;
addnewitem(Id_i, Ids_u, r = NULL)A method that adds a new item that has been consumed by already existing users in the recommendation system. Id_i: a character, the new item ID; Ids_u: a character vector, the IDs of the users who consumed the new item; r: a numeric vector, the ratings given by the users for the new item (only for ratings datatype). To add more than one new item, lists can be used. Id_i: a list of characters; Ids_u: a list of characters vectors; r: list of numeric vectors.
addnewuser(Id_u, Ids_i, r = NULL)A method that adds a new user who consumed items already existing in the recommendation system. Id_u: a character, the new user ID; Ids_i: a character vector, the IDs of the items consumed by the user; r: a numeric vector, the ratings of the items consumed by the new user (only for ratings datatype). To add more than one new user, lists can be used. Id_u: a list of characters; Ids_i: a list of characters vectors; r: list of numeric vectors.
changerating(Id_u, Id_i, r = NULL)A method that changes a rating or consumption of a user for an item that has already been rated by them. Id_u: a character, the user ID; Id_i: a character, the item ID; r: a numeric, the rating given by Id_u for Id_i (only for ratings datatype). To change more than one ratings, lists can be used. Id_u: a list of characters; Id_i: a list of characters; r: list of numeric vectors.
deleteitem(Id_i)A method that deletes an item from the recommendation system. Id_i: a character, the item ID; To delete more than one item, lists can be used. Id_i: a list of characters;
deleterating(Id_u, Id_i)A method that deletes a existing rating or consumption of a user for an item. Id_u: a character, the user ID; Id_i: a character, the item ID; To deletes more than one ratings, lists can be used. Id_u: a list of characters; Id_i: a list of characters.
deleteuser(Id_u)A method that deletes an user from the recommendation system. Id_u: a character, the user ID; To delete more than one user, lists can be used. Id_u: a list of characters;
newrating(Id_u, Id_i, r = NULL)A method that adds a new rating or consumption of an existing user for an existing item that had not yet been rated by them. Id_u: a character, the user ID; Id_i: a character, the item ID; r: a numeric, the rating given by Id_u for Id_i (only for ratings datatype). To add more than one new ratings, lists can be used. Id_u: a list of characters; Id_i: a list of characters; r: list of numeric vectors.
Jessica Kubrusly
LINDEN, G.; SMITH, B.; YORK, J. Amazon. com recommendations: Item-to-item collaborative filtering. Internet Computing, IEEE, v. 7, n. 1, p. 76-80,2003
Aggarwal, C. C. (2016). Recommender systems (Vol. 1). Cham: Springer International Publishing.
Leskovec, J., Rajaraman, A., & Ullman, J. D. (2020). Mining of massive data sets. Cambridge university press.
objectCF_r <- CFbuilder(Data = movies[1:500,], Datatype = "ratings", similarity = "pearson") dim(objectCF_r$MU) colnames(objectCF_r$MU) #movies Id rownames(objectCF_r$MU) #users Id dim(objectCF_r$SU) dim(objectCF_r$SI) objectCF_r$averages_u hist(objectCF_r$averages_u) objectCF_r$averages_i hist(objectCF_r$averages_i) objectCF_r$n_aval_u summary(objectCF_r$n_aval_u) barplot(table(objectCF_r$n_aval_u)) objectCF_r$n_aval_i summary(objectCF_r$n_aval_i) barplot(table(objectCF_r$n_aval_i)) objectCF_r$addnewuser(Id_u = "newuser1", Ids_i = "The Hunger Games: Catching Fire", r = 5) rownames(objectCF_r$MU) #users Id objectCF_r$n_aval_u["newuser1"] objectCF_r$averages_u["newuser1"] objectCF_r$addnewuser(Id_u = "newuser2", Ids_i = c("Frozen","Her","Iron Man 3"),r = c(2,4,3)) rownames(objectCF_r$MU) #users Id objectCF_r$n_aval_u["newuser2"] objectCF_r$averages_u["newuser2"] objectCF_r$addnewuser(Id_u = list("newuser3","newuser4"), Ids_i = list(c("Lincoln","Monsters University","The Lego Movie","Frozen"), c("The Wolverine","The Lego Movie")),r = list(c(1,4,5,4),c(4,5))) rownames(objectCF_r$MU) #users Id objectCF_r$n_aval_u[c("newuser3","newuser4")] objectCF_r$averages_u[c("newuser3","newuser4")] objectCF_r$newrating(Id_u = list("newuser1","newuser1","newuser2","newuser4"), Id_i = list("The Lego Movie","Wreck-It Ralph","Fast & Furious 6", "12 Years a Slave"),r = list(4,5,4,2)) objectCF_r$n_aval_u[c("newuser1","newuser2","newuser3","newuser4")] objectCF_r$averages_u[c("newuser1","newuser2","newuser3","newuser4")] objectCF_r$addnewitem(Id_i = "Oppenheimer", Ids_u = c("newuser1","newuser2","newuser3","newuser4","1","2","4","6","10", "11","20","32"),r = c(1,2,3,1,5,4,5,4,1,3,5,4)) colnames(objectCF_r$MU) objectCF_r$n_aval_i["Oppenheimer"] objectCF_r$averages_i["Oppenheimer"] objectCF_c <- CFbuilder(Data = movies[1:500,-3], Datatype = "consumption", similarity = "jaccard") dim(objectCF_c$MU) colnames(objectCF_c$MU) #movies Id rownames(objectCF_c$MU) #users Id dim(objectCF_c$SU) dim(objectCF_c$SI) objectCF_c$averages_u objectCF_c$averages_i objectCF_c$n_aval_u summary(objectCF_c$n_aval_u) barplot(table(objectCF_c$n_aval_u)) objectCF_c$n_aval_i summary(objectCF_c$n_aval_i) barplot(table(objectCF_c$n_aval_i)) objectCF_c$addnewuser(Id_u = "newuser1", Ids_i = "The Hunger Games: Catching Fire") rownames(objectCF_c$MU) #users Id objectCF_c$n_aval_u["newuser1"] objectCF_c$addnewuser(Id_u = "newuser2", Ids_i = c("Frozen","Her","Iron Man 3")) rownames(objectCF_c$MU) #users Id objectCF_c$n_aval_u["newuser2"] objectCF_c$addnewuser(Id_u = list("newuser3","newuser4"),Ids_i = list( c("Lincoln","Monsters University","The Lego Movie","Frozen"), c("The Wolverine","The Lego Movie"))) rownames(objectCF_c$MU) objectCF_c$n_aval_u[c("newuser3","newuser4")] objectCF_c$MU["newuser1","The Lego Movie"] objectCF_c$newrating(Id_u = list("newuser1","newuser1","newuser2","newuser4"), Id_i = list("The Lego Movie","Wreck-It Ralph","Fast & Furious 6", "12 Years a Slave")) objectCF_c$n_aval_u[c("newuser1","newuser2","newuser3","newuser4")] objectCF_c$averages_u[c("newuser1","newuser2","newuser3","newuser4")] objectCF_c$addnewitem(Id_i = "Oppenheimer", Ids_u = c("newuser1","newuser2","newuser3","newuser4","1","2","4","6","10", "11","20","32"),r = c(1,2,3,1,5,4,5,4,1,3,5,4)) colnames(objectCF_c$MU) objectCF_c$n_aval_i["Oppenheimer"] objectCF_c$averages_i["Oppenheimer"]objectCF_r <- CFbuilder(Data = movies[1:500,], Datatype = "ratings", similarity = "pearson") dim(objectCF_r$MU) colnames(objectCF_r$MU) #movies Id rownames(objectCF_r$MU) #users Id dim(objectCF_r$SU) dim(objectCF_r$SI) objectCF_r$averages_u hist(objectCF_r$averages_u) objectCF_r$averages_i hist(objectCF_r$averages_i) objectCF_r$n_aval_u summary(objectCF_r$n_aval_u) barplot(table(objectCF_r$n_aval_u)) objectCF_r$n_aval_i summary(objectCF_r$n_aval_i) barplot(table(objectCF_r$n_aval_i)) objectCF_r$addnewuser(Id_u = "newuser1", Ids_i = "The Hunger Games: Catching Fire", r = 5) rownames(objectCF_r$MU) #users Id objectCF_r$n_aval_u["newuser1"] objectCF_r$averages_u["newuser1"] objectCF_r$addnewuser(Id_u = "newuser2", Ids_i = c("Frozen","Her","Iron Man 3"),r = c(2,4,3)) rownames(objectCF_r$MU) #users Id objectCF_r$n_aval_u["newuser2"] objectCF_r$averages_u["newuser2"] objectCF_r$addnewuser(Id_u = list("newuser3","newuser4"), Ids_i = list(c("Lincoln","Monsters University","The Lego Movie","Frozen"), c("The Wolverine","The Lego Movie")),r = list(c(1,4,5,4),c(4,5))) rownames(objectCF_r$MU) #users Id objectCF_r$n_aval_u[c("newuser3","newuser4")] objectCF_r$averages_u[c("newuser3","newuser4")] objectCF_r$newrating(Id_u = list("newuser1","newuser1","newuser2","newuser4"), Id_i = list("The Lego Movie","Wreck-It Ralph","Fast & Furious 6", "12 Years a Slave"),r = list(4,5,4,2)) objectCF_r$n_aval_u[c("newuser1","newuser2","newuser3","newuser4")] objectCF_r$averages_u[c("newuser1","newuser2","newuser3","newuser4")] objectCF_r$addnewitem(Id_i = "Oppenheimer", Ids_u = c("newuser1","newuser2","newuser3","newuser4","1","2","4","6","10", "11","20","32"),r = c(1,2,3,1,5,4,5,4,1,3,5,4)) colnames(objectCF_r$MU) objectCF_r$n_aval_i["Oppenheimer"] objectCF_r$averages_i["Oppenheimer"] objectCF_c <- CFbuilder(Data = movies[1:500,-3], Datatype = "consumption", similarity = "jaccard") dim(objectCF_c$MU) colnames(objectCF_c$MU) #movies Id rownames(objectCF_c$MU) #users Id dim(objectCF_c$SU) dim(objectCF_c$SI) objectCF_c$averages_u objectCF_c$averages_i objectCF_c$n_aval_u summary(objectCF_c$n_aval_u) barplot(table(objectCF_c$n_aval_u)) objectCF_c$n_aval_i summary(objectCF_c$n_aval_i) barplot(table(objectCF_c$n_aval_i)) objectCF_c$addnewuser(Id_u = "newuser1", Ids_i = "The Hunger Games: Catching Fire") rownames(objectCF_c$MU) #users Id objectCF_c$n_aval_u["newuser1"] objectCF_c$addnewuser(Id_u = "newuser2", Ids_i = c("Frozen","Her","Iron Man 3")) rownames(objectCF_c$MU) #users Id objectCF_c$n_aval_u["newuser2"] objectCF_c$addnewuser(Id_u = list("newuser3","newuser4"),Ids_i = list( c("Lincoln","Monsters University","The Lego Movie","Frozen"), c("The Wolverine","The Lego Movie"))) rownames(objectCF_c$MU) objectCF_c$n_aval_u[c("newuser3","newuser4")] objectCF_c$MU["newuser1","The Lego Movie"] objectCF_c$newrating(Id_u = list("newuser1","newuser1","newuser2","newuser4"), Id_i = list("The Lego Movie","Wreck-It Ralph","Fast & Furious 6", "12 Years a Slave")) objectCF_c$n_aval_u[c("newuser1","newuser2","newuser3","newuser4")] objectCF_c$averages_u[c("newuser1","newuser2","newuser3","newuser4")] objectCF_c$addnewitem(Id_i = "Oppenheimer", Ids_u = c("newuser1","newuser2","newuser3","newuser4","1","2","4","6","10", "11","20","32"),r = c(1,2,3,1,5,4,5,4,1,3,5,4)) colnames(objectCF_c$MU) objectCF_c$n_aval_i["Oppenheimer"] objectCF_c$averages_i["Oppenheimer"]
The constructor function of the CFilt class.
CFbuilder(Data,Datatype,similarity) CFbuilder( Data, Datatype = ifelse(ncol(Data)==2,"consumption","ratings"), similarity = ifelse(Datatype == "consumption","jaccard","pearson") )CFbuilder(Data,Datatype,similarity) CFbuilder( Data, Datatype = ifelse(ncol(Data)==2,"consumption","ratings"), similarity = ifelse(Datatype == "consumption","jaccard","pearson") )
Data |
a dataframe with 2 or 3 columns. The first column indicates the user ID, the second the item ID and the third the rating (only if Datatype = 'rating'). |
Datatype |
a character that indicates the data type: 'rating' or 'consumption'. |
similarity |
a character that indicates the similarity type. For 'datatype='ratings', 'cossine' or 'person'. For datatype='consumption', 'jaccard'. |
a CF class object.
Jessica Kubrusly
LINDEN, G.; SMITH, B.; YORK, J. Amazon. com recommendations: Item-to-item collaborative filtering. Internet Computing, IEEE, v. 7, n. 1, p. 76-80,2003
CF1 <- CFbuilder(Data = movies[1:300,], Datatype = "ratings", similarity = "pearson") #or CF1_ <- CFbuilder(Data = movies[1:300,]) CF2 <- CFbuilder(Data = movies[1:300,], Datatype = "ratings", similarity = "cosine") #or CF2_ <- CFbuilder(Data = movies[1:300,], similarity = "cosine") CF3 <- CFbuilder(Data = movies[1:300,-3], Datatype = "consumption", similarity = "jaccard") #or CF3_ <- CFbuilder(Data = movies[1:300,-3])CF1 <- CFbuilder(Data = movies[1:300,], Datatype = "ratings", similarity = "pearson") #or CF1_ <- CFbuilder(Data = movies[1:300,]) CF2 <- CFbuilder(Data = movies[1:300,], Datatype = "ratings", similarity = "cosine") #or CF2_ <- CFbuilder(Data = movies[1:300,], similarity = "cosine") CF3 <- CFbuilder(Data = movies[1:300,-3], Datatype = "consumption", similarity = "jaccard") #or CF3_ <- CFbuilder(Data = movies[1:300,-3])
Functions that returns the cosine similarity between two items or users.
cosine(CF, type, i, j)cosine(CF, type, i, j)
CF |
A CF objec |
type |
"user" or "item" |
i |
"user" or "item" Id or index |
j |
"user" or "item" Id or index#' |
Jessica Kubrusly
objectCF_r <- CFbuilder(Data = movies[1:500,], Datatype = "ratings", similarity = "cosine") cosine(CF=objectCF_r,type = "user",i="1",j="2") cosine(CF=objectCF_r,type = "item",i="Her",j="Frozen")objectCF_r <- CFbuilder(Data = movies[1:500,], Datatype = "ratings", similarity = "cosine") cosine(CF=objectCF_r,type = "user",i="1",j="2") cosine(CF=objectCF_r,type = "item",i="Her",j="Frozen")
Function that provide an estimate of the user's rating for the item.
estimaterating( CF, Id_u, Id_i, type = "user", neighbors = ifelse(type == "user", nrow(CF$MU) - 1, ncol(CF$MU) - 1) )estimaterating( CF, Id_u, Id_i, type = "user", neighbors = ifelse(type == "user", nrow(CF$MU) - 1, ncol(CF$MU) - 1) )
CF |
A CF object |
Id_u |
the user Id |
Id_i |
the item Id |
type |
"user" or "item" |
neighbors |
number of neighbors in the calculation. |
Jessica Kubrusly
objectCF_r <- CFbuilder(Data = movies[1:500,], Datatype = "ratings", similarity = "cosine") estimaterating(CF=objectCF_r,Id_u="35",Id_i="Despicable Me 2") estimaterating(CF=objectCF_r,Id_u="35",Id_i="Her")objectCF_r <- CFbuilder(Data = movies[1:500,], Datatype = "ratings", similarity = "cosine") estimaterating(CF=objectCF_r,Id_u="35",Id_i="Despicable Me 2") estimaterating(CF=objectCF_r,Id_u="35",Id_i="Her")
Functions that returns the Jaccard similarity between two items or users.
jaccard(CF, type, i, j)jaccard(CF, type, i, j)
CF |
A CF objec |
type |
"user" or "item" |
i |
"user" or "item" Id or index |
j |
"user" or "item" Id or index#' |
Jessica Kubrusly
objectCF_r <- CFbuilder(Data = movies[1:500,c(1,2)], Datatype = "consumption", similarity = "jaccard") jaccard(CF=objectCF_r,type = "user",i="1",j="2") jaccard(CF=objectCF_r,type = "item",i="Her",j="Frozen")objectCF_r <- CFbuilder(Data = movies[1:500,c(1,2)], Datatype = "consumption", similarity = "jaccard") jaccard(CF=objectCF_r,type = "user",i="1",j="2") jaccard(CF=objectCF_r,type = "item",i="Her",j="Frozen")
Functions that provide items to be recommended to system users.
kclosestitems(CF, Id_i, k = 10)kclosestitems(CF, Id_i, k = 10)
CF |
A CF objec |
Id_i |
the item Id |
k |
an integer |
Jessica Kubrusly
objectCF_r <- CFbuilder(Data = movies[1:500,], Datatype = "ratings", similarity = "pearson") kclosestitems(CF = objectCF_r, Id_i = "The Lego Movie") kclosestitems(CF = objectCF_r, Id_i = "Lincoln", k=5)objectCF_r <- CFbuilder(Data = movies[1:500,], Datatype = "ratings", similarity = "pearson") kclosestitems(CF = objectCF_r, Id_i = "The Lego Movie") kclosestitems(CF = objectCF_r, Id_i = "Lincoln", k=5)
A dataset containing 7276 ratings for 50 movies by 526 users. This database was created by Giglio (2014).
moviesmovies
A data frame with 7276 rows and 3 variables:
Users identifier. Numbers 1 to 526.
Movies identifier. Movies list:
Iron Man 3
Despicable Me 2
My Mom Is a Character
Fast & Furious 6
The Wolverine
Thor: The Dark World
Hansel & Gretel: Witch Hunters
Wreck-It Ralph
Monsters University
The Hangover Part III
Vai Que Dá Certo
Meu Passado me Condena
We’re So Young
Brazilian Western
O Concurso
Mato sem Cachorro
Cine Holliudy
Odeio o Dia dos Namorados
Argo
Django Unchained
Life of Pi
Lincoln
Zero Dark Thirty
Les Miserables
Silver Linings Playbook
Beasts of the Southern Wild
Amour
A Royal Affair
American Hustle
Capitain Phillips
12 Years a Slave
Dallas Buyers Club
Gravity
Her
Philomena
The Wolf of Wall Street
The Hunt
Frozen
Till Luck Do Us Part 2
Muita Calma Nessa Hora 2
Paranormal Activity: The Marked Ones
I, Frankenstein,
The Legend of Tarzan
The Book Thief
The Lego Movie, , ,
Walking With Dinosaurs
The Hunger Games: Catching Fire
Blue Is The Warmest Color
Reaching for the Moon
The Hobbit: The Desolation of Smaug
Movie ratings by users. The ratings follows the Likert scale: 1 to 5.
Functions that returns the pearson similarity between two items or users.
pearson(CF, type, i, j)pearson(CF, type, i, j)
CF |
A CF objec |
type |
"user" or "item" |
i |
"user" or "item" Id or index |
j |
"user" or "item" Id or index#' |
Jessica Kubrusly
objectCF_r <- CFbuilder(Data = movies[1:500,], Datatype = "ratings", similarity = "pearson") pearson(CF=objectCF_r,type = "user",i="2",j="3") pearson(CF=objectCF_r,type = "item",i="Her",j="Frozen")objectCF_r <- CFbuilder(Data = movies[1:500,], Datatype = "ratings", similarity = "pearson") pearson(CF=objectCF_r,type = "user",i="2",j="3") pearson(CF=objectCF_r,type = "item",i="Her",j="Frozen")
Functions that provide the top k items to be recommended to the user Id_u.
topkitems(CF, Id_u, k = 10, type = "user")topkitems(CF, Id_u, k = 10, type = "user")
CF |
A CF objec |
Id_u |
the user Id |
k |
an integer |
type |
"user" or "item" |
Jessica Kubrusly
objectCF_r <- CFbuilder(Data = movies[1:500,], Datatype = "ratings", similarity = "pearson") u1 = rownames(objectCF_r$MU)[1] topkitems(CF=objectCF_r,Id_u = u1) u2 = rownames(objectCF_r$MU)[2] topkitems(CF=objectCF_r,Id_u = u2)objectCF_r <- CFbuilder(Data = movies[1:500,], Datatype = "ratings", similarity = "pearson") u1 = rownames(objectCF_r$MU)[1] topkitems(CF=objectCF_r,Id_u = u1) u2 = rownames(objectCF_r$MU)[2] topkitems(CF=objectCF_r,Id_u = u2)
Functions that provide the top k users to recommend the item Id_i.
topkusers(CF, Id_i, k = 10, type = "user")topkusers(CF, Id_i, k = 10, type = "user")
CF |
A CF objec |
Id_i |
the item Id |
k |
an integer |
type |
"user" or "item" |
Jessica Kubrusly
objectCF_r <- CFbuilder(Data = movies[1:500,], Datatype = "ratings", similarity = "pearson") colnames(objectCF_r$MU) topkusers(CF = objectCF_r, Id_i = "The Lego Movie") topkusers(CF = objectCF_r, Id_i = "Her")objectCF_r <- CFbuilder(Data = movies[1:500,], Datatype = "ratings", similarity = "pearson") colnames(objectCF_r$MU) topkusers(CF = objectCF_r, Id_i = "The Lego Movie") topkusers(CF = objectCF_r, Id_i = "Her")