def relim(rinput, min_support=2):
'''Finds frequent item sets of items appearing in a list of transactions
based on Recursive Elimination algorithm by Christian Borgelt.
In my synthetic tests, Relim outperforms other algorithms by a large
margin. This is unexpected as FP-Growth is supposed to be superior, but
this may be due to my implementation of these algorithms.
:param rinput: The input of the algorithm. Must come from
`get_relim_input`.
:param min_support: The minimal support of a set to be included.
:rtype: A set containing the frequent item sets and their support.
'''
fis = set()
report = {}
_relim(rinput, fis, report, min_support)
return report
def relim(rinput, min_support=2):
'''Finds frequent item sets of items appearing in a list of transactions
based on Recursive Elimination algorithm by Christian Borgelt.
In my synthetic tests, Relim outperforms other algorithms by a large
margin. This is unexpected as FP-Growth is supposed to be superior, but
this may be due to my implementation of these algorithms.
:param rinput: The input of the algorithm. Must come from
`get_relim_input`.
:param min_support: The minimal support of a set to be included.
:rtype: A set containing the frequent item sets and their support.
'''
fis = set()
report = {}
_relim(rinput, fis, report, min_support)
return report
def test_relim(should_print=False, ts=None, support=2):
if ts is None:
ts = get_default_transactions()
relim_input = get_relim_input(ts, lambda e: e)
fis = set()
report = {}
n = _relim(relim_input, fis, report, support)
if should_print:
print(n)
print(report)
return (n, report)
def test_relim(should_print=False, ts=None, support=2):
if ts is None:
ts = get_default_transactions()
relim_input = get_relim_input(ts, lambda e: e)
fis = set()
report = {}
n = _relim(relim_input, fis, report, support)
if should_print:
print(n)
print(report)
return (n, report)
def _relim(rinput, fis, report, min_support):
(relim_input, key_map) = rinput
n = 0
# Maybe this one isn't necessary
#a = deque(relim_input)
a = relim_input
while len(a) > 0:
item = a[-1][0][1]
s = a[-1][0][0]
if s >= min_support:
fis.add(item[1])
#print('Report {0} with support {1}'.format(fis, s))
report[frozenset(fis)] = s
b = _new_relim_input(len(a) - 1, key_map)
rest_lists = a[-1][1]
for (count, rest) in rest_lists:
if not rest:
continue
k = rest[0]
index = key_map[k]
new_rest = rest[1:]
# Only add this rest if it's not empty!
((k_count, k), lists) = b[index]
if len(new_rest) > 0:
lists.append((count, new_rest))
b[index] = ((k_count + count, k), lists)
n = n + 1 + _relim((b, key_map), fis, report, min_support)
fis.remove(item[1])
rest_lists = a[-1][1]
for (count, rest) in rest_lists:
if not rest:
continue
k = rest[0]
index = key_map[k]
new_rest = rest[1:]
((k_count, k), lists) = a[index]
if len(new_rest) > 0:
lists.append((count, new_rest))
a[index] = ((k_count + count, k), lists)
a.pop()
return n
def _relim(rinput, fis, report, min_support):
(relim_input, key_map) = rinput
n = 0
# Maybe this one isn't necessary
#a = deque(relim_input)
a = relim_input
while len(a) > 0:
item = a[-1][0][1]
s = a[-1][0][0]
if s >= min_support:
fis.add(item[1])
#print('Report {0} with support {1}'.format(fis, s))
report[frozenset(fis)] = s
b = _new_relim_input(len(a) - 1, key_map)
rest_lists = a[-1][1]
for (count, rest) in rest_lists:
if not rest:
continue
k = rest[0]
index = key_map[k]
new_rest = rest[1:]
# Only add this rest if it's not empty!
((k_count, k), lists) = b[index]
if len(new_rest) > 0:
lists.append((count, new_rest))
b[index] = ((k_count + count, k), lists)
n = n + 1 + _relim((b, key_map), fis, report, min_support)
fis.remove(item[1])
rest_lists = a[-1][1]
for (count, rest) in rest_lists:
if not rest:
continue
k = rest[0]
index = key_map[k]
new_rest = rest[1:]
((k_count, k), lists) = a[index]
if len(new_rest) > 0:
lists.append((count, new_rest))
a[index] = ((k_count + count, k), lists)
a.pop()
return n
