def __init__(self, db, idb=None, movies=None, k=5, sim_method=sim_cosine,
r_method='uniform', rec_type='ub'):
self.k = k
self.movies = movies
self.sim_method = sim_method
if rec_type == 'ub':
self.db = db # a dictionary for all items.
self.idb = idb
else:
self.db = idb
self.idb = db
self.r_method = r_method
self.adapt = MovieLensAdapter(self.db, self.idb) # adapter
self.parallel = get_p_dist()
class BaseEstimator(object):
def __init__(self, db, idb=None, movies=None, k=5, sim_method=sim_cosine,
r_method='uniform', rec_type='ub'):
self.k = k
self.movies = movies
self.sim_method = sim_method
if rec_type == 'ub':
self.db = db # a dictionary for all items.
self.idb = idb
else:
self.db = idb
self.idb = db
self.r_method = r_method
self.adapt = MovieLensAdapter(self.db, self.idb) # adapter
self.parallel = get_p_dist()
def find_neighbors(self, e, item):
if self.idb:
try:
potential_users = dict(self.idb[item])
# There is no potential user in training data who watch this movie
except KeyError:
return None # It will be handled by "find_distances" method
if e in potential_users.keys():
potential_users.pop(e)
return potential_users
else:
#TODO: If we haven`t any different idb dictionary..
pass
def find_distances(self, e, potential_users):
#mov_e = self.adapt.convertMovie2array(e)
n_list = []
for p_user, rating in potential_users.iteritems():
key = str(set([e, p_user]))
#if key in self.parallel:
#print "parallelized"
#return self.parallel[key]
e_A, p_user_A = self.adapt.sim_prep(e, p_user,
self.sim_method.func_name)
# If there is no item in common, no need to evaluate the distance
#if e_A.any():
#dist = self.sim_method(e_A, p_user_A, signature=key)
#n_list.append([p_user, dist, rating])
#if e_A is not None:
#dist = self.sim_method(e_A, p_user_A, signature=key)
#n_list.append([p_user, dist, rating])
try:
dist = self.sim_method(e_A, p_user_A, signature=key)
n_list.append([p_user, dist, rating])
except:
pass
n_list = np.array(n_list)
I = np.argsort(n_list[:, 1]) # sorting by distance
n_neighbors = n_list[I, :]
try:
k = n_neighbors[:self.k, :]
except IndexError:
print 'IndexError'
exit()
return k
