def k_distance(dataset, k, sd_away, distance=proximity.euclidean):
''' compute the k-distance as radius for dbscan.
k-distance is obtained by the min. distance immediately
larger than a particular standard deviation away from mean '''
global cached
if not cached:
proximity.build_cache(dataset, distance)
cached = True
kdist = [pairs[k][1] for pairs in [proximity.cache[a] for a in dataset]]
kdist.sort()
size = len(kdist)
mean = sum(kdist) / float(size)
diff = [(d - mean)**2 for d in kdist]
sd = math.sqrt(sum(diff) / float(size - 1))
anchor = mean + (sd * sd_away)
print 'mean:', mean, 'sd:', sd, 'sd + mean:', anchor,
last = None
for d in reversed(kdist):
if d > anchor:
last = d
if d < anchor and last is not None:
print 'kdist:', last
return last
raise Exception('sd too far away from mean')
def k_distance(dataset, k, sd_away, distance=proximity.euclidean):
''' compute the k-distance as radius for dbscan.
k-distance is obtained by the min. distance immediately
larger than a particular standard deviation away from mean '''
global cached
if not cached:
proximity.build_cache(dataset, distance)
cached = True
kdist = [
pairs[k][1] for pairs in
[proximity.cache[a] for a in dataset]
]
kdist.sort()
size = len(kdist)
mean = sum(kdist) / float(size)
diff = [(d - mean) ** 2 for d in kdist]
sd = math.sqrt(sum(diff) / float(size - 1))
anchor = mean + (sd * sd_away)
print 'mean:', mean, 'sd:', sd, 'sd + mean:', anchor,
last = None
for d in reversed(kdist):
if d > anchor:
last = d
if d < anchor and last is not None:
print 'kdist:', last
return last
raise Exception('sd too far away from mean')
def find_neighbour(instance, dataset, radius, distance):
''' find all neighbour within radius '''
global cached
if not cached:
proximity.build_cache(dataset, distance)
cached = True
pairs = proximity.cache[instance]
neighbour = [which for which, dist in pairs if radius >= dist]
return neighbour
def find_neighbour(instance, dataset, radius, distance):
''' find all neighbour within radius '''
global cached
if not cached:
proximity.build_cache(dataset, distance)
cached = True
pairs = proximity.cache[instance]
neighbour = [which for which, dist in pairs if radius >= dist]
return neighbour
