Preparing transaction data

Michael Hahsler

Data for association rule mining comes from many sources and in several layouts. arules stores these data in the sparse transactions class, and the transactions() constructor accepts several common input layouts.

The following examples show how to convert each layout. Always inspect the resulting transactions object with summary() or itemLabels(): values that were encoded incorrectly in the source data may otherwise become unintended items.

From a list of baskets

Use one character vector per transaction. List names become transaction IDs.

baskets <- list(
  order_1 = c("apple", "bread"),
  order_2 = c("bread", "milk"),
  order_3 = c("apple", "bread", "milk")
)
from_list <- transactions(baskets)
inspect(from_list)
#>     items                transactionID
#> [1] {apple, bread}       order_1      
#> [2] {bread, milk}        order_2      
#> [3] {apple, bread, milk} order_3

Check both the transaction summary and the resulting item labels.

summary(from_list)
#> transactions as itemMatrix in sparse format with
#>  3 rows (elements/itemsets/transactions) and
#>  3 columns (items) and a density of 0.7777778 
#> 
#> most frequent items:
#>   bread   apple    milk (Other) 
#>       3       2       2       0 
#> 
#> element (itemset/transaction) length distribution:
#> sizes
#> 2 3 
#> 2 1 
#> 
#>    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
#>   2.000   2.000   2.000   2.333   2.500   3.000 
#> 
#> includes extended item information - examples:
#>   labels
#> 1  apple
#> 2  bread
#> 3   milk
#> 
#> includes extended transaction information - examples:
#>   transactionID
#> 1       order_1
#> 2       order_2
#> 3       order_3
itemLabels(from_list)
#> [1] "apple" "bread" "milk"

The item labels confirm that the baskets were translated correctly.

From a binary matrix

Rows represent transactions and columns represent items. Logical matrices make the intended coding explicit.

binary <- matrix(
  c(TRUE, TRUE, FALSE,
    FALSE, TRUE, TRUE,
    TRUE, TRUE, TRUE),
  nrow = 3,
  byrow = TRUE,
  dimnames = list(names(baskets), c("apple", "bread", "milk"))
)
from_matrix <- transactions(binary)

itemLabels(from_matrix)
#> [1] "apple" "bread" "milk"
inspect(from_matrix)
#>     items                transactionID
#> [1] {apple, bread}       order_1      
#> [2] {bread, milk}        order_2      
#> [3] {apple, bread, milk} order_3

From a data frame in wide format

Categorical columns are converted to items of the form variable=value. Logical columns represent the presence or absence of a single item. Missing values are omitted.

customers <- data.frame(
  age_group = factor(c("young", "adult", "adult")),
  region = factor(c("north", "south", "north")),
  subscriber = c(TRUE, FALSE, TRUE)
)
from_wide <- transactions(customers)

itemLabels(from_wide)
#> [1] "age_group=adult" "age_group=young" "region=north"    "region=south"   
#> [5] "subscriber"
inspect(from_wide)
#>     items                                       transactionID
#> [1] {age_group=young, region=north, subscriber} 1            
#> [2] {age_group=adult, region=south}             2            
#> [3] {age_group=adult, region=north, subscriber} 3

Continuous variables need to be discretized before conversion.

measurements <- data.frame(
  spend = c(12, 18, 35, 42, 55),
  visits = c(1, 2, 3, 5, 8)
)
measurements_discrete <- discretizeDF(
  measurements,
  default = list(method = "frequency", breaks = 2)
)
from_discrete <- transactions(measurements_discrete)

itemLabels(from_discrete)
#> [1] "spend=[12,35)" "spend=[35,55]" "visits=[1,3)"  "visits=[3,8]"
inspect(from_discrete)
#>     items                         transactionID
#> [1] {spend=[12,35), visits=[1,3)} 1            
#> [2] {spend=[12,35), visits=[1,3)} 2            
#> [3] {spend=[35,55], visits=[3,8]} 3            
#> [4] {spend=[35,55], visits=[3,8]} 4            
#> [5] {spend=[35,55], visits=[3,8]} 5

From a data frame in long format

Long-format data has one row per transaction–item pair. Identify the transaction and item columns with cols.

long <- data.frame(
  order = c(1, 1, 2, 2, 3),
  product = c("apple", "bread", "bread", "milk", "apple")
)
from_long <- transactions(long, format = "long", cols = c("order", "product"))

itemLabels(from_long)
#> [1] "apple" "bread" "milk"
inspect(from_long)
#>     items          transactionID
#> [1] {apple, bread} 1            
#> [2] {bread, milk}  2            
#> [3] {apple}        3