def _formCluster(index, data_set):
data = data_set[index]
x = data.get_x()
y = data.get_y()
z = data.get_z()
cluster = Cluster.ClusterCenter(x, y, z, x, y, z, data.get_class())
return cluster
def _refactorClusters(clusters):
# change so when x_pos is updated the original location is not lost
new_clusters = []
for cluster in clusters:
num_of_data_points = 1
x_total = float(cluster.get_x())
y_total = float(cluster.get_y())
z_total = float(cluster.get_z())
for data_point in cluster.get_list():
x_total += float(data_point.get_x())
y_total += float(data_point.get_y())
z_total += float(data_point.get_z())
num_of_data_points += 1
x_average = x_total / num_of_data_points
y_average = y_total / num_of_data_points
z_average = z_total / num_of_data_points
new_clusters.append(
Cluster.ClusterCenter(x_average, y_average, z_average,
cluster.get_original_x(),
cluster.get_original_y(),
cluster.get_original_z(),
cluster.get_class()))
num_of_changes = 0
for i in range(3):
for data_point in clusters[i].get_list():
closest_cluster = 0
closest_distance = sys.maxsize
for j in range(3):
euclidean_distance = math.sqrt(
math.pow(
float(data_point.get_x()) -
new_clusters[j].get_x(), 2) + math.pow(
float(data_point.get_y()) -
new_clusters[j].get_y(), 2) + math.pow(
float(data_point.get_z()) -
new_clusters[j].get_z(), 2))
if euclidean_distance < closest_distance:
closest_distance = euclidean_distance
closest_cluster = j
new_clusters[closest_cluster].add_to_list(data_point)
if i != closest_cluster:
num_of_changes += 1
if num_of_changes >= 3:
return _refactorClusters(new_clusters)
else:
return new_clusters
