class PSOCKM(object):
def __init__(self, n_clusters=2, swarm_size=100, n_iter=500, w=0.72, c1=1.49, c2=1.49):
self.n_clusters = n_clusters
self.swarm_size = swarm_size
self.n_iter = n_iter
self.isTrained = False
self.up_w = w
self.lb_w = w
self.c1 = c1
self.c2 = c2
self.v_max = 0.5
self.pso = None
# self.debugger = PartitionalClusteringDebugger()
def fit(self, data):
self.isTrained = True
# self.debugger = PartitionalClusteringDebugger()
self.n_attributes = data.shape[1]
initializer = PSOCKM_SearchSpaceInitializer(self.n_clusters, data)
# initializer = UniformSSInitializer()
self.pso = PSO(PSOCKMObjectiveFunction(data, self.n_clusters, self.n_attributes),
search_space_initializer=initializer,
swarm_size=self.swarm_size,
n_iter=self.n_iter,
lb_w=self.lb_w, c1=self.c1, c2=self.c2, v_max=self.v_max)
self.pso.optimize()
self.centroids = {}
raw_centroids = self.pso.gbest_particle.pos.reshape((self.n_clusters, self.n_attributes))
for c in range(len(raw_centroids)):
self.centroids[c] = raw_centroids[c]
kmeans = KMeans(n_clusters=self.n_clusters, n_iter=1000)
kmeans.set_initial_solution(dict(self.centroids))
kmeans.fit(data)
self.centroids = dict(kmeans.centroids)
self.solution = ClusteringSolution(centroids=self.centroids, dataset=data)
return self.n_iter
def reset(self):
self.isTrained = False
def predict(self, x):
if self.isTrained:
dist = []
for c in self.centroids:
temp_centroid = np.tile(self.centroids[c], (np.shape(x)[0], 1))
diff = (x - self.centroids[c])
power2 = (x - self.centroids[c]) ** 2
sum = np.sum((x - self.centroids[c]) ** 2, axis=1)
dist.append(sum)
labels = np.argmin(dist, axis=0)
return np.array(labels).T
# return np.argmin(np.array(class_).T, axis=1)
raise Exception("NonTrainedModelException: You must fit data first!")
class PSOKM(object):
def __init__(self,
n_clusters=2,
swarm_size=100,
n_iter=500,
w=0.72,
lb_w=0.4,
w_damp=None,
c1=1.49,
c2=1.49):
self.n_clusters = n_clusters
self.swarm_size = swarm_size
self.n_iter = n_iter
self.w = w
self.lb_w = lb_w
self.c1 = c1
self.c2 = c2
self.run = False
self.v_max = None
if w_damp is None:
self.w_damp = self.w - self.lb_w
def fit(self, data):
self.run = True
self.dim = data.shape[1]
self.pso = PSO(dim=self.dim * self.n_clusters,
minf=0,
maxf=1,
swarm_size=self.swarm_size,
n_iter=self.n_iter,
w=self.w,
lb_w=self.lb_w,
c1=self.c1,
c2=self.c2)
self.pso.optimize(self.__objective_function,
customizable=True,
dim=self.dim,
n_clusters=self.n_clusters,
data=data)
self.centroids = {}
raw_centroids = self.pso.global_optimum.pos.reshape(
(self.n_clusters, self.dim))
for centroid in range(len(raw_centroids)):
self.centroids[centroid] = raw_centroids[centroid]
kmeans = KMeans(n_clusters=self.n_clusters)
kmeans.set_initial_solution(dict(self.centroids))
kmeans.fit(data)
self.centroids = dict(kmeans.centroids)
def predict(self, x):
if self.run:
if len(x.shape) > 1:
class_ = []
for c in self.centroids:
class_.append(
np.sum((x - self.centroids[c].best_post)**2, axis=1))
return np.argmin(np.array(class_).T, axis=1)
else:
dist = [
np.linalg.norm(x - self.centroids[c])
for c in self.centroids
]
class_ = dist.index(min(dist))
return class_
else:
raise Exception(
"NonTrainedModelException: You must fit data first!")
def __objective_function(self, particle, **kwargs):
if ('dim' not in kwargs) or ('n_clusters' not in kwargs) or (
('data' not in kwargs)):
raise Exception(
'Illegal Arguments Exception: Expected arguments does not match!'
)
dim = kwargs.get("dim")
data = kwargs.get("data")
n_clusters = kwargs.get("n_clusters")
centroids = particle.reshape((n_clusters, dim))
clusters = {}
for k in range(n_clusters):
clusters[k] = []
for xi in data:
dist = [
np.linalg.norm(xi - centroids[c])
for c in range(len(centroids))
]
class_ = dist.index(min(dist))
clusters[class_].append(xi)
inter_cluster_sum = 0.0
for c in range(len(centroids)):
intra_sum = 0.0
if len(clusters[c]) > 0:
for point in clusters[c]:
intra_sum += np.linalg.norm(point - centroids[c])
intra_sum = intra_sum / len(clusters[c])
inter_cluster_sum += intra_sum
inter_cluster_sum = inter_cluster_sum / len(centroids)
return inter_cluster_sum