def map_compute_distance(args):
if len(args) == 4:
dist_matrix, dicts, mode, ctgy_chkin_dict = args
queue = None
else:
dist_matrix, queue, dicts, mode, ctgy_chkin_dict = args
inner_dist = list()
outer_dist = list()
for i, dist in enumerate(dist_matrix):
if mode == 'sub':
same_ctgy_chkins = ctgy_chkin_dict[dicts.chkin_ctgy_dict[
dicts.reverse_dictionary[i]]]
else:
same_ctgy_chkins = ctgy_chkin_dict[
dicts.ctgy_mapping['subctgy_ctgy_dict'][correct_errata(
dicts.chkin_ctgy_dict[dicts.reverse_dictionary[i]])]]
ids = [
dicts.dictionary[key] for key in same_ctgy_chkins
if key in dicts.dictionary.keys()
]
same_ctgy_mask = np.ones(dist_matrix.shape[1]).astype(int)
same_ctgy_mask[ids] = 0
diff_ctgy_mask = np.logical_not(same_ctgy_mask)
same_dists = ma.masked_array(dist, mask=same_ctgy_mask)
diff_dists = ma.masked_array(dist, mask=diff_ctgy_mask)
inner_dist.append(np.mean(same_dists))
outer_dist.append(np.mean(diff_dists))
if not queue is None:
queue.put((inner_dist, outer_dist))
else:
return inner_dist, outer_dist
def get_labels(dicts):
subctgys = sorted(dicts.ctgy_chkin_dict.keys())
subctgy_dictionary = {k:i for i, k in enumerate(subctgys)}
sublabels = [subctgy_dictionary[dicts.chkin_ctgy_dict[dicts.reverse_dictionary[i]]] for i in range(dicts.vocabulary_size-1)]
rootctgys = sorted(dicts.rootctgy_chkin_dict.keys())
rootctgy_dictionary = {k:i for i, k in enumerate(rootctgys)}
rootlabels = [rootctgy_dictionary[dicts.ctgy_mapping['subctgy_ctgy_dict'][correct_errata(dicts.chkin_ctgy_dict[dicts.reverse_dictionary[i]])]] for i in range(dicts.vocabulary_size-1)]
return {'sub':sublabels, 'root':rootlabels}
def get_same_ctgy_chkins(i, dicts, mode):
if mode == 'root':
ctgy_chkin_dict = dicts.rootctgy_chkin_dict
same_ctgy_chkins = ctgy_chkin_dict[dicts.ctgy_mapping['subctgy_ctgy_dict'][
correct_errata(dicts.chkin_ctgy_dict[dicts.reverse_dictionary[i]]) ]]
else:
ctgy_chkin_dict = dicts.ctgy_chkin_dict
same_ctgy_chkins = ctgy_chkin_dict[dicts.chkin_ctgy_dict[dicts.reverse_dictionary[i]]]
return same_ctgy_chkins
def get_ctgy_histogram_dict(traj_list, dicts):
timebins = list(range(24))
substats = {k:np.zeros_like(timebins) for k in dicts.ctgy_chkin_dict}
rootstats = {k:np.zeros_like(timebins) for k in dicts.ctgy_mapping['ctgy_subctgy_dict']}
for seq in traj_list:
for p in seq:
poi = p[0]
if dicts.reverse_dictionary[poi] == 'UNK':
continue
sctgy = dicts.chkin_ctgy_dict[dicts.reverse_dictionary[poi]]
rctgy = dicts.ctgy_mapping['subctgy_ctgy_dict'][correct_errata(sctgy)]
visit_time = p[-1].hour
substats[sctgy][visit_time] += 1
rootstats[rctgy][visit_time] += 1
return substats, rootstats
def get_rootctgy_chkin(ctgy_chkin_dict, subctgy_ctgy_dict):
newdict = defaultdict(list)
for key in ctgy_chkin_dict:
newdict[subctgy_ctgy_dict[correct_errata(key)]].extend(
ctgy_chkin_dict[key])
return newdict
def id2rootctgy(dicts, ctgy_mapping, idx):
return ctgy_mapping['subctgy_ctgy_dict'][correct_errata(dicts.chkin_ctgy_dict[dicts.reverse_dictionary[idx]])]