def centroid_and_links(nodes, n_clusters):
clusters, cluster_series = spatial.zone_clusters(
nodes,
n_clusters=n_clusters,
geo_union_method=lambda lg: shapely.geometry.MultiPoint(list(lg)),
geo_join_method=geo_join_method)
index_name = cluster_series.index.name
index_name = index_name if index_name else 'index'
grouped = cluster_series.reset_index().groupby('cluster')
first = list(grouped[index_name].first())
node_lists = list(grouped[index_name].agg(lambda s: list(s)))
node_geo_dict = nodes['geometry'].to_dict()
def link_geometry(a, b):
return shapely.geometry.LineString(
[node_geo_dict[a], node_geo_dict[b]])
values = []
for node_list in node_lists:
for a in node_list:
for b in node_list:
if a != b:
values.append([a, b, link_geometry(a, b)])
links = pd.DataFrame(values, columns=['a', 'b', 'geometry'])
links['length'] = skims.distance_from_geometry(links['geometry'])
return first, links
def renumber(zones,
volume,
n_clusters=10,
volume_columns=['volume'],
cluster_column=None):
clusters, cluster_series = spatial.zone_clusters(zones, n_clusters,
cluster_column)
grouped = renumber_volume(volume,
cluster_series,
volume_columns=volume_columns)
return clusters, grouped, cluster_series
def skim_matrix(zones,
token,
n_clusters,
coordinates_unit='degree',
skim_matrix_kwargs={}):
clusters, cluster_series = spatial.zone_clusters(zones, n_clusters, 1e-9)
cluster_euclidean = all_skim_matrix(
clusters, token, coordinates_unit=coordinates_unit**skim_matrix_kwargs)
df = euclidean(zones, coordinates_unit=coordinates_unit)
df = pd.merge(df,
pd.DataFrame(cluster_series),
left_on='origin',
right_index=True)
df = pd.merge(df,
pd.DataFrame(cluster_series),
left_on='destination',
right_index=True,
suffixes=['_origin', '_destination'])
df = pd.merge(df,
cluster_euclidean.rename(
columns={
'origin': 'cluster_origin',
'destination': 'cluster_destination',
'distance': 'cluster_distance',
'duration': 'cluster_duration'
}),
on=['cluster_origin', 'cluster_destination'],
suffixes=['', '_cluster'])
df['distance_rate'] = (df['euclidean_distance'] /
df['euclidean_distance_cluster']).fillna(0)
df['distance'] = df['cluster_distance'] * df['distance_rate']
df['duration'] = df['cluster_duration'] * df['distance_rate']
euclidean_to_path_length = 1 / (df['euclidean_distance_cluster'] /
df['cluster_distance']).mean()
euclidean_speed = (df['euclidean_distance_cluster'] /
df['duration']).mean()
df.loc[df['euclidean_distance_cluster'] == 0,
'duration'] = df['euclidean_distance'] / euclidean_speed
df.loc[df['euclidean_distance_cluster'] == 0,
'distance'] = df['euclidean_distance'] * euclidean_to_path_length
return df.fillna(0)
def renumber_quetzal(zones,
volume,
od_stack,
n_clusters=10,
cluster_column=None,
volume_columns=['volume'],
volume_od_columns=['volume_pt'],
distance_columns=['euclidean_distance']):
clusters, cluster_series = spatial.zone_clusters(
zones, n_clusters=n_clusters, cluster_column=cluster_column)
grouped = renumber_volume(volume,
cluster_series,
volume_columns=volume_columns)
od_stack_grouped = renumber_od_stack(od_stack, cluster_series,
volume_od_columns, distance_columns)
return clusters, grouped, cluster_series, od_stack_grouped
def node_clustering(links,
nodes,
n_clusters,
prefixe='',
group_id=None,
**kwargs):
disaggregated_nodes = nodes.copy()
if group_id is None:
clusters, cluster_series = spatial.zone_clusters(nodes,
n_clusters=n_clusters,
**kwargs)
else:
clusters = nodes.groupby(group_id).first()
cluster_series = nodes[group_id]
cluster_dict = cluster_series.to_dict()
centroids = clusters.copy()
centroids['geometry'] = centroids['geometry'].apply(lambda g: g.centroid)
try:
links = links.copy()
links['disaggregated_a'] = links['a']
links['disaggregated_b'] = links['b']
links['a'] = links['a'].apply(lambda x: prefixe + str(cluster_dict[x]))
links['b'] = links['b'].apply(lambda x: prefixe + str(cluster_dict[x]))
except AttributeError:
links = None
clusters['count'] = cluster_series.value_counts()
disaggregated_nodes['cluster'] = cluster_series
parenthood = pd.merge(disaggregated_nodes,
centroids,
left_on='cluster',
right_index=True,
suffixes=['_node', '_centroid'])
parenthood['geometry'] = parenthood.apply(parenthood_geometry, axis=1)
centroids.index = prefixe + pd.Series(centroids.index).astype(str)
return links, centroids, clusters, parenthood
def cluster_snail_number(zones, n_clusters=20, centre=None, buffer=10):
"""
zones: GeoSeries
"""
df = pd.DataFrame(zones).reset_index().copy()
if centre is None:
union = cascaded_union(df.geometry).buffer(buffer)
centre = union.centroid
# Snail clusterize
clusters, cluster_series = spatial.zone_clusters(df, n_clusters=n_clusters)
df['cluster'] = cluster_series
snail = snail_number(clusters, centre)
clusters['snail'] = snail
df = df.merge(snail.reset_index(), on='cluster')
df.drop('cluster', 1, inplace=True)
# snail numbering within cluster
to_concat = []
for cluster in set(df['cluster_snail']):
temp_df = df.loc[df['cluster_snail']==cluster]
temp_centre = cascaded_union(temp_df.geometry).centroid
temp_snail = snail_number(temp_df, temp_centre)
temp_df['snail'] = temp_snail
to_concat.append(temp_df)
concat = pd.concat(to_concat)
concat = concat.sort_values(['cluster_snail', 'snail']).reset_index(drop=True)
concat = concat.reset_index().rename(
columns={
'level_0': 'id',
'cluster_snail': 'cluster',
'index': 'original_index'
}
)
return concat[['cluster', 'id', 'original_index']]