index.r

library("arules")


# Exercise 1

# Import csv and create a unique set of all the grocery names
groceries <- read.csv("./GroceriesInitial.csv", header=TRUE, sep=",")
product_names = unique(unlist(groceries[,4,35]))

# Transform groceries to a binary correlation values
products <- as.data.frame(t(apply(groceries[,4:35], 1, function(x)
    (product_names) %in% as.character(unlist(x)))))
names(products) <- product_names

# Add `id`, `basket_value` and `recency_days` columns
groceries <- data.frame(cbind(groceries[,1:3], products))

to_keep <- c(
    "id",
    "basket_value",
    "recency_days",
    "citrus.fruit",
    "tropical.fruit",
    "whole.milk",
    "other.vegetables",
    #"rolls/buns",
    "chocolate",
    "bottled.water",
    "yogurt",
    "sausage",
    "root.vegetables",
    "pastry",
    "soda",
    "cream"
)

# filter columns and keep only the wanted ones
groceries <- groceries[ , to_keep]

# Create 3 new columns with default value FALSE
groceries$low_value_basket <- FALSE
groceries$medium_value_basket <- FALSE
groceries$high_value_basket <-  FALSE

# Set the values of these columns conditionaly to true
groceries$low_value_basket[groceries$basket_value <= 2.5] <- TRUE
groceries$medium_value_basket[groceries$basket_value > 2.5 & groceries$basket_value <= 6] <- TRUE
groceries$high_value_basket[groceries$basket_value > 6] <- TRUE

# The 3 counts below are roughly the same
str(sum(groceries$low_value_basket == TRUE))
str(sum(groceries$medium_value_basket == TRUE))
str(sum(groceries$high_value_basket == TRUE))


# Exercise 2

only_items <- groceries[,4:ncol(groceries)]
only_items$low_value_basket <- NULL
only_items$medium_value_basket <- NULL
only_items$high_value_basket <- NULL

rules <- apriori(
    only_items,
    parameter = list(minlen=2, maxlen=3, supp = 0.01, conf = 0.5),
    control = list(verbose = FALSE)
)

rules_sorted_only_items <- sort(rules, by="lift")
#print(inspect(rules_sorted_only_items))

rules <- apriori(
    groceries[,1:ncol(groceries)],
    parameter = list(minlen=2, maxlen=3, supp = 0.01, conf = 0.5),
    control = list(verbose = FALSE)
)

rules_sorted_all <- sort((rules), by="lift")
#print(inspect(rules_sorted_all))