def Cluster(self):
# read in the coords within the safe range
coords = self.openfile()
# set radians for the haversine metric
kms_per_radian = 6371.0088
# set epsilon to 50 meters
epsilon = 0.05 / kms_per_radian
# set epslion really small so it forms a cluster on each plot
espsi = 0.0000000000000000001 / kms_per_radian
# perform the DBSCAN using the ball tree algorithm and haversine metric.
# form cluster if 10 pics are present within 50 meters
db = DBSCAN(eps=epsilon, min_samples=10, algorithm='ball_tree', metric='haversine').fit(np.radians(coords))
# form cluster on every single plot
scat = DBSCAN(eps=espsi, min_samples=1, algorithm='ball_tree', metric='haversine').fit(np.radians(coords))
cluster_labels = db.labels_
scatcluster_labels = scat.labels_
# set number of clusters to the number of labels (one per cluster)
num_clusters = len(set(cluster_labels))
if num_clusters < 3:
print('Not enough clusters were formed')
PlotCluster()
scatnum_clusters = len(set(scatcluster_labels))
print('Number of clusters: {}'.format(scatnum_clusters))
# convert the coords to a DataFrame with columns of lat and lon
scat = pd.DataFrame(coords, columns=["latitude", "longitude"])
df = pd.DataFrame(coords, columns=["latitude", "longitude"])
# convert to matrix to allow manipulation into series
scatcoords = scat.as_matrix(columns=["latitude", "longitude"])
coords = df.as_matrix(columns=["latitude", "longitude"])
# store clusters in a series data type
clusters = pd.Series([coords[cluster_labels == n] for n in range(num_clusters - 1)])
scatclusters = pd.Series([scatcoords[scatcluster_labels == n] for n in range(scatnum_clusters - 1)])
def get_centermost_point(cluster):
# using the centroid attribute, pull out the centre point of each cluster
centroid = (MultiPoint(cluster).centroid.x, MultiPoint(cluster).centroid.y)
# save the centermost point in a variable and return as a tuple
centermost_point = min(cluster, key=lambda point: great_circle(point, centroid).m)
print(centermost_point)
return tuple(centermost_point)
# use series' map attribute to map the centermost point to each cluster
centermost_points = clusters.map(get_centermost_point)
# also map every point to the clusters of each point to allow for plotting
scatplot = scatclusters.map(get_centermost_point)
# separate the lat and lon points from the tuple
Alatitude, Alongitude = zip(*scatplot)
if len(centermost_points) > 2:
latitude, longitude = zip(*centermost_points)
else:
print("Not enough data to form clusters")
# plot the centermost points using matplotlib
plt.scatter(latitude, longitude, s=2, c="black", marker=".")
plt.show()
# use gmap to open the map to the first plot point
gmap = gmplot.GoogleMapPlotter(longitude[0], latitude[0], zoom=15)
# plot the centermost points and every other point
gmap.scatter(longitude, latitude, '#ff0000', size=10, marker=False)
gmap.scatter(Alongitude, Alatitude, '#0000ff', size=5, marker=False)
# save map as html file in directory
gmap.draw(userlocation + 'Map.html')
