def load_model(kind, model_fdir):
"""
Get FCM model from the model file
:param kind:
:param model_fdir:
:return:
"""
df = pd.read_csv(model_fdir,
delimiter='\t',
names=['property_uri', 'kind', 'features'])
dfkind = df[df.kind == kind]
# print(dfkind.columns.values)
centroids = []
#centroids_names = list(df['property_uri'])
centroids_names = []
# centroids_names = [p for p in df['property_uri']]
for idx, row in dfkind.iterrows():
#for r in dfkind['features']:
# print(r)
# print(type(r))
r = row['features']
centroids_names.append(row['property_uri'])
centroid = [float(num) for num in r.split(',')]
centroids.append(centroid)
# This is to fix for categorical
max_num_features = 0
if kind == commons.CATEGORICAL: # has variant num of features
for c in centroids:
dim = len(c)
if dim > max_num_features:
max_num_features = dim
for c in centroids:
num_feats = len(c)
if num_feats < max_num_features:
additionals = max_num_features - num_feats
for i in range(additionals):
c.append(0)
logger.debug("for categorical max_num_features: " +
str(max_num_features))
if len(centroids) == 0:
return None, None, None
fcm = FCM(n_clusters=len(centroids))
# print("centroids: "+str(centroids))
# print("len: "+str(len(centroids)))
fcm.fit(centroids, range(len(centroids)))
logger.debug("centroids: " + str(centroids))
logger.debug("fcm centroids: " + str(fcm.cluster_centers_))
logger.debug("membership: ")
logger.debug("\n" + str(fcm.u))
return fcm, centroids_names, max_num_features
def example():
X = np.array([[1, 1], [1, 2], [2, 2], [9, 10], [10, 10], [10, 9], [9, 9],
[20, 20]])
fcm = FCM(n_clusters=3)
fcm.fit(X, [0, 0, 0, 1, 1, 1, 1, 2])
# fcm.fit(X)
testing_data = np.array([[0, 1.9], [5, 3], [4, 4], [8, 9], [9.5, 6.5],
[5, 5], [15, 15], [12, 12], [14, 14], [19, 10]])
predicted_membership = fcm.predict(testing_data)
print("\n\ntesting data")
print(testing_data)
print("predicted membership")
print(predicted_membership)
print("\n\n")
draw_model_2d(fcm, data=testing_data, membership=predicted_membership)
def test_2d():
X = np.array([[1, 1], [1, 2], [2, 2], [9, 10], [10, 10], [10, 9], [9, 9]])
fcm = FCM()
fcm.fit(X, [0, 0, 0, 1, 1, 1, 1])
assert len(fcm.cluster_centers_) == 2
assert fcm.cluster_centers_[0][0] == pytest.approx(1.33333333, 0.1)
assert fcm.cluster_centers_[0][1] == pytest.approx(1.66666667, 0.1)
testing_data = np.array([[0, 1.9], [3, 3], [4, 4], [8, 9], [9.5, 6.5]])
predicted_membership = fcm.predict(testing_data)
actual_membership = np.array([[0.98777232, 0.01222768],
[0.94884591, 0.05115409],
[0.82813688, 0.17186312],
[0.02482074, 0.97517926],
[0.0908581, 0.9091419]])
assert predicted_membership == pytest.approx(actual_membership, 0.01)
def example():
X = np.array([[1, 1], [1, 2], [2, 2], [9, 10], [10, 10], [10, 9], [9, 9],
[20, 20]])
fcm = FCM(n_clusters=3)
fcm.set_logger(tostdout=True, level=logging.DEBUG)
fcm.fit(X, [0, 0, 0, 1, 1, 1, 1, 2])
# fcm.fit(X)
testing_data = np.array([[0, 1.9], [5, 3], [4, 4], [8, 9], [9.5, 6.5],
[5, 5], [15, 15], [12, 12], [14, 14], [19, 10]])
predicted_membership = fcm.predict(testing_data)
print "\n\ntesting data"
print testing_data
print "predicted membership"
print predicted_membership
print "\n\n"
draw_model_2d(fcm, data=testing_data, membership=predicted_membership)
def fit_from_centroids():
data_test = [178.75, 5.97390157]
data_height = [170., 186., 182., 177.]
data_size = [40, 42, 45, 47, 50]
data_injuries = [3, 4, 1, 0]
centers = [[175., 4.08], [44.8, 3.54], [2., 1.58]]
fcm = FCM(n_clusters=3, max_iter=1)
fcm.cluster_centers_ = np.array(centers)
fcm.fit(fcm.cluster_centers_, [0, 1, 2])
membership_height = fcm.predict(np.array([data_test]))
membership_height = fcm.predict(
np.array([[np.average(data_height),
np.std(data_height)]]))
membership_size = fcm.predict(
np.array([[np.average(data_size),
np.std(data_size)]]))
membership_injuries = fcm.predict(
np.array([[np.average(data_injuries),
np.std(data_injuries)]]))
def load_mlmodel_into_fcm(model):
"""
:param model: MLModel
:param data: matrix input as np array
:return: FCM model
"""
num_of_cols = len(model.cluster_set.all())
centers = []
for clus_model in model.cluster_set.all():
logger.debug("load_mlmodel_into_fcm> getting clus %s" %
clus_model.name)
p = clus_model.center.split(',')
centers.append(p)
centers_np = np.array(centers, dtype='f')
# logger.debug("Centers: ")
# logger.debug(centers_np)
fcm = FCM(n_clusters=num_of_cols, max_iter=1, logger=logger)
fcm.cluster_centers_ = centers_np
data = centers_np
logger.debug("load_mlmodel_into_fcm> will fit the data with %d clusters" %
num_of_cols)
fcm.fit(data, range(num_of_cols))
logger.info("generated the FCM model from the MLModel")
return fcm
def single_known_cluster():
X = np.array([[1, 1], [1, 2], [2, 2]])
fcm = FCM(n_clusters=3, max_iter=1)
fcm.fit(X, [0, 0, 0])
def example_single_known():
X = np.array([[1, 1], [1, 2], [2, 2], [0, 0], [0, 0]])
fcm = FCM(n_clusters=3, max_iter=1)
fcm.fit(X, [0, 0, 0, 1, 2])
draw_model_2d(fcm, data=X, membership=fcm.u)
print fcm.u