Data engineering and BI courses are free!
The transactions class represents transaction data used for
mining itemsets or rules.
Objects are created by coercion from objects of other classes (see Examples section) or by calls of the form
new("transactions", ...)
or by using the constructor function
transactions(x, itemLabels = NULL, transactionInfo = NULL, format = "wide", cols = NULL).
Format "wide" is a regular data.frame where each row contains an object. Format "long" is a data.frame with one column with transaction IDs and one with an item. cols is a numeric or character vector of length two giving the numbers or names of the columns (fields) with the transaction and item ids, respectively.
See Examples Section for creating transactions from data.
itemsetInfo:a data.frame
with one row per transaction (each transaction is considered an
itemset). The data.frame
can hold columns with additional information, e.g.,
transaction IDs or user IDs for each transaction. Note: this
slot is inherited from class '>itemMatrix, but
should be accessed in transactions with the
method transactionInfo().
data:object of class
ngCMatrix to store the
binary incidence matrix (see
'>itemMatrix class)
itemInfo:signature(from = "matrix", to = "transactions");
produces a transactions data set from a binary incidence matrix.
The column names are used as item labels and the row names are
stores as transaction IDs.
signature(from = "transactions", to = "matrix");
coerces the transactions data set into a binary incidence matrix.
signature(from = "list", to = "transactions");
produces a transactions data set from a list. The names of the
items in the list are used as item labels.
signature(from = "transactions", to = "list");
coerces the transactions data set into a list of transactions.
Each transaction is a vector of character strings (names of the
contained items).
signature(from = "data.frame", to = "transactions");
recodes the data frame containing only categorical variables (factors)
or logicals all into a binary transaction data set. For binary variables
only TRUE values are converted into items and the item label is the
variable name. For factors, a dummy item for each level is
automatically generated. Item labels are generated by concatenating
variable names and levels with "=".
The original variable names and levels are stored in the itemInfo
data frame
as the components variables and levels.
Note that NAs are ignored (i.e., do not generate an item).
signature(from = "transactions", to = "data.frame");
represents the set of transactions in a printable form
as a data.frame.
Note that this does not reverse coercion from data.frame
to transactions.
signature(from = "ngCMatrix", to = "transactions");
Note that the data is stored transposed in the ngCMatrix. Items are
stored as rows and transactions are columns!
signature(x = "transactions");
returns row (transactionID) and column (item) names.
signature(x = "transactions");
returns the items in the transactions as an
'>itemMatrix.
signature(x = "transactions");
returns the labels for the itemsets in each transaction
(see itemMatrix).
signature(x = "transactions");
replaces the transaction information with a new data.frame.
signature(x = "transactions");
returns the transaction information as a data.frame.
signature(object = "transactions"); convert the transactions to long format (a data.frame with two columns, tid and item). Column names can be specified as a character vector of length 2 called cols.
signature(object = "transactions")
signature(object = "transactions")
Transactions are a direct extension of class
'>itemMatrix to store a binary incidence
matrix, item labels, and optionally transaction IDs and user IDs.
Transactions can be created from a list containing transactions or a matrix or data.frames using
the constructor function transactions(x, itemLabels = NULL, transactionInfo = NULL), or
S4 coercion with as(x, "transactions")).
itemLabels and
transactionInfo are by default created from information in x (e.g., from row and column names). In the constructor function, the user can specify for itemLabels a vector of all possible item labels (character)
or another transactions object to copy the item coding (see itemCoding for details).
Note that you will need to prepare your data first (see coercion methods in the Methods Section and the Example Section below for details on the needed format).
Continuous variables: Association rule mining can only use items and does not work with continuous variables. Continuous variables need to be discretized first. An item resulting from discretization might be age>18 and the column contains only TRUE or FALSE. Alternatively it can be a factor with levels age<=18, 50=>age>18 and age>50. These will be automatically converted into 3 items, one for each level. Have a look at the function discretize for automatic discretization.
Logical variables: A logical variable describing a person could be tall indicating if the person is tall using the values TRUE and FALSE. The fact that the person is tall would be encoded in the transaction containing the item tall while not tall persons would not have this item. Therefore, for logical variables, the TRUE value is converted into an item with the name of the variable and for the FALSE values no item is created.
Factors: The function also can convert columns with nominal values (i.e., factors) into a series of binary items (one for each level constructed as `variable name`=`level`). Note that nominal variables need to be encoded as factors (and not characters or numbers). This can be done with
data[,"a_nominal_var"] <- factor(data[,"a_nominal_var"]).
Complete examples for how to prepare data can be found in the man pages for Income and
Adult.
Transactions are represented as sparse binary matrices of class
itemMatrix. If you work with several transaction sets at the
same time, then the encoding (order of the items in the binary matrix) in the different sets is important.
See itemCoding to learn how to encode and recode transaction sets.
[-methods,
discretize,
LIST,
write,
c,
image,
inspect,
itemCoding,
read.transactions,
random.transactions,
sets,
itemMatrix-class
# NOT RUN {
## Example 1: creating transactions form a list (each element is a transaction)
a_list <- list(
c("a","b","c"),
c("a","b"),
c("a","b","d"),
c("c","e"),
c("a","b","d","e")
)
## Set transaction names
names(a_list) <- paste("Tr",c(1:5), sep = "")
a_list
## Use the constructor to create transactions
## Note: S4 coercion does the same trans1 <- as(a_list, "transactions")
trans1 <- transactions(a_list)
trans1
## Analyze the transactions
summary(trans1)
image(trans1)
## Example 2: creating transactions from a 0-1 matrix with 5 transactions (rows) and
## 5 items (columns)
a_matrix <- matrix(c(
1, 1, 1, 0, 0,
1, 1, 0, 0, 0,
1, 1, 0, 1, 0,
0, 0, 1, 0, 1,
1, 1, 0, 1, 1
), ncol = 5)
## Set item names (columns) and transaction labels (rows)
colnames(a_matrix) <- c("a", "b", "c", "d", "e")
rownames(a_matrix) <- paste("Tr", c(1:5), sep = "")
a_matrix
## Create transactions
trans2 <- transactions(a_matrix)
trans2
inspect(trans2)
## Example 3: creating transactions from data.frame (wide format)
a_df <- data.frame(
age = as.factor(c(6, 8, NA, 9, 16)),
grade = as.factor(c("A", "C", "F", NA, "C")),
pass = c(TRUE, TRUE, FALSE, TRUE, TRUE))
## Note: factors are translated differently than logicals and NAs are ignored
a_df
## Create transactions
trans3 <- transactions(a_df)
inspect(trans3)
## Note that coercing the transactions back to a data.frame does not recreate the
## original data.frame, but represents the transactions as sets of items
as(trans3, "data.frame")
## Example 4: creating transactions from a data.frame with
## transaction IDs and items (long format)
a_df3 <- data.frame(
TID = c( 1, 1, 2, 2, 2, 3 ),
item = c("a", "b", "a", "b", "c", "b")
)
a_df3
trans4 <- transactions(a_df3, format = "long", cols = c("TID", "item"))
trans4
inspect(trans4)
## convert transactions back into long format.
toLongFormat(trans4)
## Example 5: create transactions from a dataset with numeric variables
## using discretization.
data(iris)
irisDisc <- discretizeDF(iris)
head(irisDisc)
trans5 <- transactions(irisDisc)
trans5
inspect(head(trans5))
## Note, creating transactions without discretizing numeric variables will apply the
## default discretization and also create a warning.
## Example 6: create transactions manually (with the same item coding as in trans5)
trans6 <- transactions(
list(
c("Sepal.Length=[4.3,5.4)", "Species=setosa"),
c("Sepal.Length=[4.3,5.4)", "Species=setosa")
), itemLabels = trans5)
trans6
inspect(trans6)
# }
Run the code above in your browser using DataLab