def test_route_steps_distance(self):
matrix_instance = DistanceMatrix(os.path.dirname(os.path.abspath(".")))
df = pd.read_csv(os.path.join(os.path.abspath(".."),
r"Data\Geocodes.csv"),
error_bad_lines=False)
new_row = pd.DataFrame({
'latitude': 51.742503,
'longitude': 19.432956
},
index=[0])
df = pd.concat([new_row, df]).reset_index(drop=True)
matrix = matrix_instance.calculate_distance_matrix(df)
self.assertEqual(matrix.shape[0], df.shape[0])
route, cost = tsp.min_distance_search(df.shape[0])
df_brews = pd.read_csv(os.path.join(os.path.abspath(".."),
r"Data\Breweries.csv"),
error_bad_lines=False)
df_beers = pd.read_csv(os.path.join(os.path.abspath(".."),
r"Data\Beers.csv"),
error_bad_lines=False)
collected, breweries = tsp.collect_beer(route, df_beers, df, df_brews)
self.assertEqual(len(breweries), len(route))
for index, brewery in enumerate(breweries):
if index != 0:
self.assertLessEqual(
abs(brewery[3] -
np.round(matrix[route[index - 1],
route[index]], 0)[0]), 1)
def __init__(self, distMatrtix, alignment=None, names=None):
self._alignment = alignment
self._names = names
if self._alignment:
self._distMatrix = DistanceMatrix(distMatrtix, [x.id for x in self._alignment])
elif self._names:
self._distMatrix = DistanceMatrix(distMatrtix, self._names)
else:
raise RuntimeError("You must pass either an alignment or a list of names.")
def test_find_nearest_with_one_circle_point(self):
distance_matrix = DistanceMatrix()
distance_matrix.add_distance(3.8, 'blue')
result = distance_matrix.find_nearest_label(k=1)
expected = ['blue']
self.assertEqual(expected, result)
def test_find_nearest_with_same_label(self):
distance_matrix = DistanceMatrix()
distance_matrix.add_distance(3.8, 'blue')
distance_matrix.add_distance(3.8, 'blue')
result = distance_matrix.find_nearest_label(k=1)
expected = ['blue']
self.assertEqual(sorted(expected), sorted(result))
def put_distance_matrices(self, metrics, tmpdir="/tmp", normalise=False):
"""
Pass this function a list of metrics
valid kwargs - invert (bool), normalise (bool)
"""
if not isinstance(metrics, list):
metrics = [metrics]
trees = [rec.tree for rec in self.get_records()]
for metric in metrics:
dm = DistanceMatrix(trees, tmpdir=tmpdir)
dm.get_distance_matrix(metric, normalise=normalise)
self.distance_matrices[metric] = dm
def test_distance_matrix(self):
df = pd.read_csv(os.path.join(os.path.abspath(".."),
r"Data\Geocodes.csv"),
error_bad_lines=False)
df = df.drop(columns=['id', 'brewery_id', 'accuracy'])
df = df.head(30)
matrix_instance = DistanceMatrix(os.path.dirname(os.path.abspath(".")))
matrix = matrix_instance.test_disttance_matrix(30)
for row_idx, row in enumerate(matrix):
for val_idx, value in enumerate(row):
actual_val = haversine.haversine(df.iloc[row_idx]['longitude'],
df.iloc[row_idx]['latitude'],
df.iloc[val_idx]['longitude'],
df.iloc[val_idx]['latitude'])
self.assertEqual(value, actual_val)
def fit(self, n_clusters):
'''
this method uses the main Divisive Analysis algorithm to do the clustering
arguements
----------
n_clusters - integer
number of clusters we want
returns
-------
cluster_labels - numpy array
an array where cluster number of a sample corrosponding to
the same index is stored
'''
similarity_matrix = DistanceMatrix(
self.data) # similarity matrix of the data
clusters = [
list(range(self.n_samples))
] # list of clusters, initially the whole dataset is a single cluster
while True:
c_diameters = [
np.max(similarity_matrix[cluster][:, cluster])
for cluster in clusters
] #cluster diameters
max_cluster_dia = np.argmax(c_diameters) #maximum cluster diameter
max_difference_index = np.argmax(
np.mean(similarity_matrix[clusters[max_cluster_dia]]
[:, clusters[max_cluster_dia]],
axis=1))
splinters = [clusters[max_cluster_dia][max_difference_index]
] #spinter group
last_clusters = clusters[max_cluster_dia]
del last_clusters[max_difference_index]
while True:
split = False
for j in range(len(last_clusters))[::-1]:
splinter_distances = similarity_matrix[last_clusters[j],
splinters]
last_distances = similarity_matrix[
last_clusters[j],
np.delete(last_clusters, j, axis=0)]
if np.mean(splinter_distances) <= np.mean(last_distances):
splinters.append(last_clusters[j])
del last_clusters[j]
split = True
break
if split == False:
break
del clusters[max_cluster_dia]
clusters.append(splinters)
clusters.append(last_clusters)
if len(clusters) == n_clusters:
break
cluster_labels = np.zeros(self.n_samples)
for i in range(len(clusters)):
cluster_labels[clusters[i]] = i
return cluster_labels
def get_distances(csv_file):
csv_file = CsvHelper(csv_file)
try:
url_index = 0
price_index = 1
address_index = 10
distance_matrix = {}
for index, row in enumerate(csv_file.read()):
if index == 0:
continue
url = row[url_index]
price = int(row[price_index].replace("£",
"").replace("$", "").replace(
"€", "").replace(",", ""))
address = row[address_index]
if price > 200000:
continue
print("distance matrix:", url)
distance_matrix[index] = DistanceMatrix(address).get()
finally:
csv_file.file.close()
return distance_matrix
def put_distance_matrices(
self,
metrics,
tmpdir='/tmp',
normalise=False,
):
"""
Pass this function a list of metrics
valid kwargs - invert (bool), normalise (bool)
"""
if not isinstance(metrics, list):
metrics = [metrics]
trees = [rec.tree for rec in self.get_records()]
for metric in metrics:
dm = DistanceMatrix(trees, tmpdir=tmpdir)
dm.get_distance_matrix(metric, normalise=normalise)
self.distance_matrices[metric] = dm
def __init__(self, dm):
if isinstance(dm, np.ndarray):
dm = DistanceMatrix(dm)
if not isinstance(dm, DistanceMatrix):
raise ValueError(
'Distance matrix should be a numpy array or treeCl.DistanceMatrix'
)
self.dm = dm
def test_indexing(y):
x = pdist(y)
s = squareform(x)
d = DistanceMatrix(x)
assert d[0, 0] == 0
assert d[0, 1] == d[1, 0]
assert d[0, 1] == s[0, 1]
assert d[1, 0] == s[1, 0]
def test_attrs(y):
x = pdist(y)
s = squareform(x)
d = DistanceMatrix(x)
assert d.T == d
# np.testing.assert_array_equal(d.flat[0], x[0])
assert d.size == s.size
assert d.ndim == 2
assert d.shape == s.shape
assert len(d) == d.shape[0]
def get_inter_tree_distances(self,
metric,
normalise=False,
batchsize=100,
background=False):
""" Generate a distance matrix from a fully-populated Collection """
array = _get_inter_tree_distances(metric, self.trees, normalise,
batchsize, background)
if background: # return IPython.parallel map result object to the user before jobs are finished
return array
return DistanceMatrix(array, self.names)
def test_find_nearest_with_two_identical_labels(self):
distance_matrix = DistanceMatrix()
distance_matrix.add_distance(3.8, 'blue')
distance_matrix.add_distance(3, 'blue')
result = distance_matrix.find_nearest_label(k=3)
expected = ['blue']
self.assertEqual(expected, result)
def test_find_nearest_point_between_two(self):
distance_matrix = DistanceMatrix()
distance_matrix.add_distance(3.8, 'red')
distance_matrix.add_distance(3, 'blue')
result = distance_matrix.find_nearest_label(k=1)
expected = ['blue']
self.assertEqual(expected, result)
def test_find_most_present_label_between_three(self):
distance_matrix = DistanceMatrix()
distance_matrix.add_distance(3.8, 'red')
distance_matrix.add_distance(3.8, 'red')
distance_matrix.add_distance(3, 'blue')
result = distance_matrix.find_nearest_label(k=3)
expected = ['red']
self.assertEqual(expected, result)
def test_find_most_present_ambigous_labels_between_four(self):
distance_matrix = DistanceMatrix()
distance_matrix.add_distance(3.8, 'red')
distance_matrix.add_distance(3.8, 'red')
distance_matrix.add_distance(3, 'blue')
distance_matrix.add_distance(3, 'blue')
result = distance_matrix.find_nearest_label(k=3)
expected = ['blue', 'red']
self.assertEqual(sorted(expected), sorted(result))
def write_distances(csv_file, json_file):
original_csv_file = CsvHelper(csv_file)
new_csv_file = CsvHelper(f"new_props_{time.time()}")
distance_matrix = JsonHelper(json_file).read()
try:
for index, row in enumerate(original_csv_file.read()):
key = str(index + 1)
if key in distance_matrix:
result = DistanceMatrix.to_h(distance_matrix[key])
price = int(row["Price"].replace("£",
"").replace("$", "").replace(
"€", "").replace(",", ""))
row["Price"] = price
address_index = 10
row_keys = list(row.keys())
row_values = list(row.values())
postal_info = Postal(row["Address"]).info()
postal_length = len(postal_info)
postal_keys = list(postal_info.keys())
for index, key in enumerate(postal_keys):
offset_index = address_index + index + 1
row_keys.insert(offset_index, key)
postal_values = list(postal_info.values())
for index, value in enumerate(postal_values):
offset_index = address_index + index + 1
row_values.insert(offset_index, value)
keys = list(result.keys())
for index, key in enumerate(keys):
offset_index = address_index + postal_length + index + 1
row_keys.insert(offset_index, key)
values = list(result.values())
for index, value in enumerate(values):
offset_index = address_index + postal_length + index + 1
row_values.insert(offset_index, value)
new_row = dict(zip(row_keys, row_values))
new_csv_file.write(new_row)
finally:
original_csv_file.file.close()
new_csv_file.file.close()
def main(source_path, destination_path, tensor_dimension):
for root, dirs, files in os.walk(source_path, topdown=False):
for name in tqdm(files):
dis = DistanceMatrix(source_path, destination_path,
name.split('_')[0], tensor_dimension)
try:
dis.standardise_flatten_distance_matrix()
except:
pass
try:
dis.save_file()
except:
pass
def test_complex(y):
x = pdist(y)
d = DistanceMatrix(x)
np.testing.assert_array_equal(d.imag.values, DistanceMatrix(x.imag).values)
np.testing.assert_array_equal(d.real.values, DistanceMatrix(x.real).values)
def test_conversions(y):
x = pdist(y)
d = DistanceMatrix(x)
s = squareform(x)
np.testing.assert_array_equal(d.toarray(), np.triu(s))
assert (d.tosparse() != sps.coo_matrix(np.triu(s))).nnz == 0
def update_distance_matrix(lat, lon):
matrix = DistanceMatrix(os.path.dirname(os.path.abspath(__file__)))
matrix.update_csv(lat, lon, os.path.dirname(os.path.abspath(__file__)))
def dist(self, obj1, obj2):
distance = DistanceMatrix([obj1.tree, obj2.tree], self.metric)[0][1]
return (-distance)
class NeighborJoining:
##
# The constructor just saves the data for the execution.
#
# @param distMatrtix the distance matrix
# @param alignment should be specified when the names parameter is not present
# @param names the names of the taxa in the distance matrix
def __init__(self, distMatrtix, alignment=None, names=None):
self._alignment = alignment
self._names = names
if self._alignment:
self._distMatrix = DistanceMatrix(distMatrtix, [x.id for x in self._alignment])
elif self._names:
self._distMatrix = DistanceMatrix(distMatrtix, self._names)
else:
raise RuntimeError("You must pass either an alignment or a list of names.")
##
# Neighbor-Joining implementation.
#
# @return constructed tree with edges weighted by distances
@property
def tree(self):
L = self._distMatrix.columnNames
tree = Tree()
tree.name = "root"
tree.dist = 0
for seq in L:
tree.add_child(name=seq, dist=0)
iter_count = 1
while len(L) > 2:
nearest_nbs = self._distMatrix.getNearestNeigbors()
node_i = tree.search_nodes(name=nearest_nbs[0])[0]
node_j = tree.search_nodes(name=nearest_nbs[1])[0]
L.remove(nearest_nbs[0])
L.remove(nearest_nbs[1])
node_k = Tree()
node_k.dist = 0
node_k.name = "X" + str(iter_count)
d_ij = self._distMatrix.getDistance(node_i.name, node_j.name)
assert d_ij > 0
d_ik = 0.5 * d_ij + 0.5 * (self._distMatrix.getSeparation(node_i.name) - self._distMatrix.getSeparation(node_j.name))
d_jk = 0.5 * d_ij + 0.5 * (self._distMatrix.getSeparation(node_j.name) - self._distMatrix.getSeparation(node_i.name))
tree.remove_child(node_i)
tree.remove_child(node_j)
node_k.add_child(node_i, dist=d_ik)
node_k.add_child(node_j, dist=d_jk)
tree.add_child(node_k)
d_km = []
for node_m in L:
d_km.append(0.5 * (self._distMatrix.getDistance(node_i.name, node_m) + self._distMatrix.getDistance(node_j.name, node_m) - d_ij) )
assert d_km > 0
self._distMatrix.removeData((node_i.name, node_j.name))
self._distMatrix.appendData(d_km, node_k.name)
iter_count+=1
L = self._distMatrix.columnNames
last_nodes = tree.get_children()
d_ij = self._distMatrix.getDistance(last_nodes[0].name, last_nodes[1].name)
leaf = None
new_root = None
for node in last_nodes:
if node.is_leaf():
node.dist = d_ij
leaf = node.detach()
else:
new_root = node.detach()
if not leaf:
leaf = last_nodes[0]
leaf.dist = d_ij
new_root.add_child(leaf)
return new_root
##
# @var _distMatrix
# the distance matrix in more or less arbitrary form
# @var _names
# taxa identification strings
# @var _alignment
# multiple sequence alignment
def test_return_only_the_minimal_necessary_point(self):
distance_matrix = DistanceMatrix()
distance_matrix.add_distance(2, 'blue')
distance_matrix.add_distance(3, 'green')
distance_matrix.add_distance(4, 'blue')
distance_matrix.add_distance(4, 'green')
distance_matrix.add_distance(4, 'red')
result = distance_matrix.find_nearest_label(k=3)
expected = ['blue', 'green']
self.assertEqual(sorted(expected), sorted(result))
def distance_matrix(self, metric, **kwargs):
""" Generate a distance matrix from a fully-populated Collection """
trees = [rec.tree for rec in self.records]
return DistanceMatrix(trees, metric, tmpdir=self.tmpdir, **kwargs)
import numpy as np
from distance_matrix import DistanceMatrix
import random
import pandas as pd
import os
matrix = DistanceMatrix(os.path.dirname(os.path.abspath(__file__)))
limit = 2000
def __initialize__(size):
random_list = random.sample(range(1, 1305), size)
route = np.array([0] + random_list + [0])
return route
def __initialize_empty__(size):
route = np.zeros((size, 2))
return route
def evaluate_distance():
pass
def min_distance_search(size):
# initialize values
route = __initialize_empty__(size) # route contains vertex id and distance to it from previous vertex
cost = 0 # total distance
vertex = 0 # current vertex
read = {} # read lines from distance matrix file (minimizes reading from file)
