def weighted_initial(self, working_set, mislabeled):
print "Total Cluster Centroids Chosen: ", len(self.mislabeled_chosen)
print len(mislabeled), " mislabeled emails remaining as possible cluster centroids"
if len(mislabeled) == 0: #No more centers to select
return (None, None, 'NO_CENTROIDS')
else:
prob, mislabeled_point = mislabeled.pop(0) # Choose most potent mislabeled email
self.mislabeled_chosen.append(mislabeled_point)
print "Chose the mislabeled point with z = ", prob
data_y, data_x = h.compose_set(working_set)
vec_data_x = vectorize_set(data_x)
init_email = None
init_pos = None
label = None
if "frequency" in self.distance_opt:
min_distance = sys.maxint
for i,email_indices in enumerate(vec_data_x):
if None not in email_indices: # actual data
current_distance = distance(email_indices, mislabeled_point, self.distance_opt)
if current_distance < min_distance:
init_email = data_x[i]
init_pos = i
min_distance = current_distance
if init_email is None:
print "Training emails remaining: ", len(data_x)
else:
label = data_y[init_pos]
print "-> selected cluster centroid with label: ", label, " and distance: ", min_distance, " from mislabeled point"
return (label, init_pos, init_email)
def __init__(self, msg, size, active_unlearner, label, distance_opt,
working_set=None, separate=True):
# Clustroid specs
self.clustroid = msg[1] # index of msg
self.label = label
self.common_features = []
self.separate = separate
self.size = size # arbitrarily set to 100
self.active_unlearner = active_unlearner # point to calling au instance
self.opt = distance_opt
# The data
self.working_set = working_set
self.train_y = self.working_set[0]
self.train_x = self.working_set[1]
self.pol_y = self.working_set[2]
self.pol_x = self.working_set[3]
self.data_y, self.data_x = h.compose_set(self.working_set)
time_1 = time.time()
self.vec_data_x = vectorize_set(self.data_x)
print 'Vectorizing data_x took: ', h.sec_to_english(time.time() - time_1)
self.ham = set()
self.spam = set()
if 'frequency' in self.opt:
self.cluster_word_frequency = msg[0] # actual vector representation of msg
self.added = [] # keeps track of order emails are added
self.dist_list = self.distance_array(self.separate) # returns list containing dist from all emails in phantom space to center clustroid
self.cluster_set = self.make_cluster() # adds closest emails to cluster
self.divide() # adds cluster emails to ham and spam
def __init__(self,
msg,
size,
active_unlearner,
label,
distance_opt,
working_set=None,
separate=True):
# Clustroid specs
self.clustroid = msg[1] # index of msg
self.label = label
self.common_features = []
self.separate = separate
self.size = size # arbitrarily set to 100
self.active_unlearner = active_unlearner # point to calling au instance
self.opt = distance_opt
# The data
self.working_set = working_set
self.train_y = self.working_set[0]
self.train_x = self.working_set[1]
self.pol_y = self.working_set[2]
self.pol_x = self.working_set[3]
self.data_y, self.data_x = h.compose_set(self.working_set)
time_1 = time.time()
self.vec_data_x = vectorize_set(self.data_x)
print 'Vectorizing data_x took: ', h.sec_to_english(time.time() -
time_1)
self.ham = set()
self.spam = set()
if 'frequency' in self.opt:
self.cluster_word_frequency = msg[
0] # actual vector representation of msg
self.added = [] # keeps track of order emails are added
self.dist_list = self.distance_array(
self.separate
) # returns list containing dist from all emails in phantom space to center clustroid
self.cluster_set = self.make_cluster(
) # adds closest emails to cluster
self.divide() # adds cluster emails to ham and spam
def weighted_initial(self, working_set, mislabeled):
print "Total Cluster Centroids Chosen: ", len(self.mislabeled_chosen)
print len(
mislabeled
), " mislabeled emails remaining as possible cluster centroids"
if len(mislabeled) == 0: #No more centers to select
return (None, None, 'NO_CENTROIDS')
else:
prob, mislabeled_point = mislabeled.pop(
0) # Choose most potent mislabeled email
self.mislabeled_chosen.append(mislabeled_point)
print "Chose the mislabeled point with z = ", prob
data_y, data_x = h.compose_set(working_set)
vec_data_x = vectorize_set(data_x)
init_email = None
init_pos = None
label = None
if "frequency" in self.distance_opt:
min_distance = sys.maxint
for i, email_indices in enumerate(vec_data_x):
if None not in email_indices: # actual data
current_distance = distance(email_indices,
mislabeled_point,
self.distance_opt)
if current_distance < min_distance:
init_email = data_x[i]
init_pos = i
min_distance = current_distance
if init_email is None:
print "Training emails remaining: ", len(data_x)
else:
label = data_y[init_pos]
print "-> selected cluster centroid with label: ", label, " and distance: ", min_distance, " from mislabeled point"
return (label, init_pos, init_email)
def cluster_remaining(au, working_set):
""" This function is called if weighted_initial returns NO_CENTROIDS, meaning there are no more misabeled emails to use as centers.
The remaining emails in the working set are then returned as one cluster.
"""
print "No more cluster centroids, grouping all remaining emails into one cluster"
first_state_rate = au.current_detection_rate
size = len(h.strip(working_set[0] + working_set[2])) # get number of remaining emails
init_email = None
init_pos = None
label = None
data_y, data_x = h.compose_set(working_set)
for i,l in enumerate(data_y): # loop to find first email that is not none
if l is not None:
label = l
init_pos = i
init_email = data_x[i]
center = (init_email, init_pos)
cluster = Cluster(center, size, au, label, au.distance_opt, working_set=working_set)
au.unlearn(cluster)
au.init_ground()
new_detection_rate = au.current_detection_rate
au.learn(cluster) # relearn cluster in real training space so deltas of future cluster are not influenced
second_state_rate = au.current_detection_rate
net_rate_change = second_state_rate - first_state_rate
au.current_detection_rate = first_state_rate
assert(au.current_detection_rate == first_state_rate), str(au.current_detection_rate) + " " + str(first_state_rate)
print "clustered remaining with a net rate change of ", second_state_rate, " - ", first_state_rate, " = ", net_rate_change
return net_rate_change, cluster
