def test_k_distance():
instances = ((1,1),(2,2),(3,3))
d = lof.k_distance(1,(2,2),instances)
assert d == (0.0, [(2,2)])
d = lof.k_distance(1,(2.2,2.2),instances)
assert d == (0.20000000000000018, [(2,2)])
d = lof.k_distance(1,(2.5,2.5),instances)
assert d == (0.5, [(2,2),(3,3)])
def cal_C3(instance):
k_inf = kinf[instances.index(instance)]
k_avg = (2 + k_inf) / 2
(k1, neighbours) = k_distance(k_inf + 1, instance, instances)
(k2, neighbours) = k_distance(k_avg, instance, instances)
c3 = k1 / k2
return c3
def test_k_distance():
instances = np.array(((1, 1), (2, 2), (3, 3)))
d = lof.k_distance(1, (2, 2), instances)
assert np.allclose(d[0], 0.0)
assert np.allclose(d[1], np.array([(2, 2)]))
d = lof.k_distance(1, (2.2, 2.2), instances)
assert np.allclose(d[0], 0.28284271247461928)
assert np.allclose(d[1], np.array([(2, 2)]))
d = lof.k_distance(1, (2.5, 2.5), instances)
assert np.allclose(d[0], 0.70710678118654757)
assert np.allclose(d[1], np.array([(2, 2), (3, 3)]))
d = lof.k_distance(5, (2.2, 2.2), instances)
assert np.allclose(d[0], 1.6970562748477143)
assert np.allclose(d[1], np.array([(2, 2), (3, 3), (1, 1)]))
def update():
# 计算邻接矩阵M
for instance in instances:
instances_value_backup = list(instances)
instances_value_backup.remove(instance)
(k_distance_value,
neighbours) = k_distance(len(instances_value_backup), instance,
instances_value_backup)
M[instances.index(instance)] = neighbours
