def test_InvertedIndex():
data = ("a,b,c,d,e,f\n"
"g,h,i,j,k,l\n"
"z,x\n"
"z,x\n"
"z,x,y\n"
"z,x,y,i\n")
index = InvertedIndex()
index.load(data)
assert (index.support({Item("a")}) == 1 / 6)
assert (index.support({Item("b")}) == 1 / 6)
assert (index.support({Item("c")}) == 1 / 6)
assert (index.support({Item("d")}) == 1 / 6)
assert (index.support({Item("e")}) == 1 / 6)
assert (index.support({Item("f")}) == 1 / 6)
assert (index.support({Item("h")}) == 1 / 6)
assert (index.support({Item("i")}) == 2 / 6)
assert (index.support({Item("j")}) == 1 / 6)
assert (index.support({Item("k")}) == 1 / 6)
assert (index.support({Item("l")}) == 1 / 6)
assert (index.support({Item("z")}) == 4 / 6)
assert (index.support({Item("x")}) == 4 / 6)
assert (index.support({Item("y")}) == 2 / 6)
sup_zx = index.support({Item("z"), Item("x")})
assert (sup_zx == 4 / 6)
sup_zxy = index.support({Item("z"), Item("x"), Item("y")})
assert (sup_zxy == 2 / 6)
sup_zxyi = index.support({Item("z"), Item("x"), Item("y"), Item("i")})
assert (sup_zxyi == 1 / 6)
def test_apriori():
data = ("a,b,c,d,e,f\n"
"g,h,i,j,k,l\n"
"z,x\n"
"z,x\n"
"z,x,y\n"
"z,x,y,i\n")
expectedItemSets = {
ItemSet("i"): 2 / 6,
ItemSet("z"): 4 / 6,
ItemSet("x"): 4 / 6,
ItemSet("y"): 2 / 6,
ItemSet("xz"): 4 / 6,
ItemSet("yz"): 2 / 6,
ItemSet("xy"): 2 / 6,
ItemSet("xyz"): 2 / 6
}
index = InvertedIndex()
index.load(data)
itemsets = apriori(index, 2 / 6)
assert (set(expectedItemSets.keys()) == set(itemsets))
for itemset in itemsets:
assert (expectedItemSets[itemset] == index.support(itemset))
print("Itemsets={}".format([i for i in itemsets if len(i) > 1]))
# (antecedent, consequent, confidence, lift, support)
expectedRules = {
(frozenset({Item("x"),
Item("y")}), frozenset({Item("z")}), 1, 1.5, 1 / 3),
(frozenset({Item("x")}), frozenset({Item("y")}), 0.5, 1.5, 1 / 3),
(frozenset({Item("x")}), frozenset({Item("z"),
Item("y")}), 0.5, 1.5, 1 / 3),
(frozenset({Item("x")}), frozenset({Item("z")}), 1, 1.5, 2 / 3),
(frozenset({Item("y")}), frozenset({Item("x")}), 1, 1.5, 1 / 3),
(frozenset({Item("y")}), frozenset({Item("z"),
Item("x")}), 1, 1.5, 1 / 3),
(frozenset({Item("y")}), frozenset({Item("z")}), 1, 1.5, 1 / 3),
(frozenset({Item("z"),
Item("x")}), frozenset({Item("y")}), 0.5, 1.5, 1 / 3),
(frozenset({Item("z"),
Item("y")}), frozenset({Item("x")}), 1, 1.5, 1 / 3),
(frozenset({Item("z")}), frozenset({Item("x"),
Item("y")}), 0.5, 1.5, 1 / 3),
(frozenset({Item("z")}), frozenset({Item("x")}), 1, 1.5, 2 / 3),
(frozenset({Item("z")}), frozenset({Item("y")}), 0.5, 1.5, 1 / 3),
}
rules = set(generate_rules(itemsets, 0, 0, index))
for (antecedent, consequent, confidence, lift, support) in rules:
print("{}, {} conf={:.4f}, {:.4f}, {:.4f}".format(
antecedent, consequent, confidence, lift, support))
assert (rules == expectedRules)
def test_apriori():
data = ("a,b,c,d,e,f\n"
"g,h,i,j,k,l\n"
"z,x\n"
"z,x\n"
"z,x,y\n"
"z,x,y,i\n")
expectedItemSets = {ItemSet("i"): 2 / 6,
ItemSet("z"): 4 / 6,
ItemSet("x"): 4 / 6,
ItemSet("y"): 2 / 6,
ItemSet("xz"): 4 / 6,
ItemSet("yz"): 2 / 6,
ItemSet("xy"): 2 / 6,
ItemSet("xyz"): 2 / 6}
index = InvertedIndex()
index.load(data)
itemsets = apriori(index, 2 / 6)
assert(len(itemsets) == len(expectedItemSets))
for itemset in itemsets:
assert(frozenset(itemset) in expectedItemSets)
for itemset in itemsets:
assert(expectedItemSets[frozenset(itemset)] == index.support(itemset))
print("Itemsets={}".format([i for i in itemsets if len(i) > 1]))
def itemize(a):
return list(map(item_id, a))
# (antecedent, consequent, confidence, lift, support)
rx = [
(['y'], ['x'], 1.0, 1.5, 0.3333333333333333),
(['x'], ['y'], 0.5, 1.5, 0.3333333333333333),
(['y'], ['z'], 1.0, 1.5, 0.3333333333333333),
(['z'], ['y'], 0.5, 1.5, 0.3333333333333333),
(['x'], ['z'], 1.0, 1.5, 0.6666666666666666),
(['z'], ['x'], 1.0, 1.5, 0.6666666666666666),
(['x', 'y'], ['z'], 1.0, 1.5, 0.3333333333333333),
(['z', 'y'], ['x'], 1.0, 1.5, 0.3333333333333333),
(['z', 'x'], ['y'], 0.5, 1.5, 0.3333333333333333),
(['y'], ['z', 'x'], 1.0, 1.5, 0.3333333333333333),
(['x'], ['z', 'y'], 0.5, 1.5, 0.3333333333333333),
(['z'], ['x', 'y'], 0.5, 1.5, 0.3333333333333333)
]
expectedRules = list(map(lambda a: (itemize(a[0]), itemize(a[1]), a[2], a[3], a[4]), rx))
itemset_counts = dict(map(lambda i: (tuple(i), index.count(i)), itemsets))
rules = generate_rules(
itemsets,
itemset_counts,
index.num_transactions,
0,
0)
def deitemize(a):
return list(map(item_str, a))
p = list(map(lambda a: (deitemize(a[0]), deitemize(a[1]), a[2], a[3], a[4]), rules))
print("rules")
print(p)
for (antecedent,
consequent,
confidence,
lift,
support) in rules:
print("{}, {} conf={:.4f}, {:.4f}, {:.4f}".
format(antecedent, consequent, confidence, lift, support))
assert(len(rules) == len(expectedRules))
for i in range(len(rules)):
assert(expectedRules[i] in rules)