tests/testthat/test.complimentarity.R

In byapparov/recommender: Provides functions for collaborative filtering

library(data.table)
context("Post order functions.")

test_that("Calculation of post order affinities", {

  # transactions is a data table that stores the history of transactions per customer.
  # UID is the identifier of a customer
  # reference is the transaction reference
  # type can be whatever grouping one wishes to impose on any sold item
  transactions <- data.table(
                  UID =       c("A",  "B",  "C",  "D",  "A",  "A",  "B",  "C"),
                  reference = c("A1", "B1", "C1", "D1", "A2", "A2", "B2", "C2"),
                  type =      c("a",  "b",  "a",  "d",  "b",  "c",  "d",  "c"))

  map <-  getProductGroupAffinities(transactions)

  expect_true(map[type == "a"]$b == 1, "Correct post order mapping info found")
  expect_true(map[type == "a"]$c == 2, "Correct post order mapping info found")

  transactions <- data.table(
    UID =       c("A",  "B",  "A",  "A" , "A",  "B",  "A"),
    reference = c("A1", "B1", "A2", "A2", "A2", "B2", "A3"),
    type =      c("a",  "b",  "b",  "c",  "d",  "a",  "a"))

  map <-  getProductGroupAffinities(transactions)

  expect_true(map[type == "a"]$b == 1, "Correct post order mapping info found")
  expect_true(map[type == "b"]$a == 2, "Correct post order mapping info found")

})

test_that("Top types are correctly retrieved", {

  # The type of the first transaction
  type <- "A"

  # Column type holds the types
  # Each of the columns holds the information about the affinity for the
  # next transaction per row. Larger numbers means better matching.
  # i.e for type A
  #     A    B   C   D
  # A   1.0  0.1 0.2 0.6
  data <- data.table(type = c("A","B","C","D"),
                     A = c(1.0, 0.4, 0.5, 0.1),
                     B = c(0.1, 1.0, 0.2, 0.8),
                     C = c(0.2, 0.6, 1.0, 0.9),
                     D = c(0.6, 0.2, 0.8, 1.0))

  res <- getNextOrderTopTypes(type, data, 1, TRUE)
  expect_equal(res, "D", "Return the correct match.")

  res <- getNextOrderTopTypes(type, data, 1, FALSE)
  expect_equal(res, "A", "Return the correct match if not excluding itself.")

  type <- c("A", "D")
  res <- getNextOrderTopTypes(type, data, 1, TRUE)
  expect_equal(res, "C", "Return the correct match multiple types.")

  type <- c("A", "D")
  res <- getNextOrderTopTypes(type, data, 1, FALSE)
  expect_equal(res, "D", "Return the correctmatch multiple types if not excluding itself.")

  type <- c("A", "D", "Error")
  res <- getNextOrderTopTypes(type, data, 1, FALSE)
  expect_equal(res, "D", "Return the correctmatch multiple types if not excluding itself
               with extra unknown columns.")

  type <- c("As", "Ds")
  res <- getNextOrderTopTypes(type, data, 1, FALSE)
  expect_equal(res, NA, "Return NA if the types are unknown.")

})


test_that("Post order recommendations correctly retrieved", {

  # Grouping assingment between skus and types/categories/subcategories
  sku.details <- data.table(sku = c("A", "B", "C", "D"),
                            category = c("A1", "A1", "A2", "A3"))

  # SKU to SKU similarity matrix
  sim.matrix <- c(1.00, 0.52, 0.41, 0.52,
                  0.60, 0.60, 0.23, 0.26,
                  0.40, 0.30, 0.83, 0.26,
                  0.62, 0.80, 0.25, 0.16 )
  sim.matrix <- matrix(sim.matrix, nrow = 4, ncol = 4)
  colnames(sim.matrix) <- rownames(sim.matrix) <- c("A", "B", "C", "D")

  # Data.table similar to the one in the test "Top types are correctly retrieved"
  post.order.matrix <- data.table(type = c("A1", "A2", "A3"),
                                  A1 = c(1.0, 0.5, 0.2),
                                  A2 = c(0.5, 1.0, 0.1),
                                  A3 = c(0.2, 0.1, 1.0))

  # Last transactions SKUS named with their type/category/subcategory
  skus <-c("A1" = "A")

  res <- getComplimentaryProducts(sim.matrix = sim.matrix,
                                  skus = skus,
                                  values = 2,
                                  groups = sku.details$category,
                                  affinity = post.order.matrix,
                                  sku.details = sku.details,
                                  n.of.types = 2)

  expect_equal(as.character(res), c("D", "C"), "Bring the correct post recommendations.")

  skus <- c("A3" = "D",
            "A2" = "C")

  res <- getComplimentaryProducts(sim.matrix = sim.matrix,
                                  skus = skus,
                                  values = 1,
                                  groups = sku.details$category,
                                  affinity = post.order.matrix,
                                  sku.details = sku.details,
                                  n.of.types = 2)

  expect_equal(as.character(res), c("A"), "Bring the correct post recommendations.")

})