def prepare_dataset(dataset_file, dataset_name, type):
dataset = pd.read_csv(resource_path + dataset_file, index_col=False)
for column in dataset.columns[0:-1]:
if column == "POCT":
triangular_set = fl.TriangularSet(constant.POCT_MIN,
constant.POCT_MEAN,
constant.POCT_MAX)
dataset[column] = triangular_set(np.array(dataset[column]))
elif column == "IGA_totali":
triangular_set = fl.TriangularSet(constant.MIN_IGA,
constant.MEAN_IGA,
constant.MAX_IGA)
dataset[column] = triangular_set(np.array(dataset[column]))
elif column == "TTG_IGG":
triangular_set = fl.TriangularSet(constant.TTG_IGG_MIN,
constant.TTG_IGG_MEAN,
constant.TTG_IGG_MAX)
dataset[column] = triangular_set(np.array(dataset[column]))
elif column == "TTG_IGA":
triangular_set = fl.TriangularSet(constant.TTG_IGA_MIN,
constant.TTG_IGA_MEAN,
constant.TTG_IGA_MAX)
dataset[column] = triangular_set(np.array(dataset[column]))
if type == ETHALONS:
dataset.to_csv(neuro_networks_path + "ethalons_" + dataset_name +
".csv",
index=False)
elif type == CANDIDATES:
dataset.to_csv(neuro_networks_path + "candidates_" + dataset_name +
".csv",
index=False)
def default_fuzzifier(idx, F):
# get min/max from data
v_min = np.nanmin(F)
v_max = np.nanmax(F)
# blarg
return [ Leaf(idx, "low", fl.TriangularSet(v_min - (v_max - v_min) ** 2, v_min, v_max)),
Leaf(idx, "med", fl.TriangularSet(v_min, v_min + ((v_max - v_min) / 2), v_max)),
Leaf(idx, "hig", fl.TriangularSet(v_min, v_max, v_max + (v_max - v_min) ** 2)) ]
def default_fuzzifier(idx, F):
"""Default fuzzifier function.
Creates three fuzzy sets with triangular membership functions: (low, med, hig) from min and max data points.
"""
# get min/max from data
v_min = np.nanmin(F)
v_max = np.nanmax(F)
# blarg
return [
Leaf(idx, "low",
fl.TriangularSet(v_min - (v_max - v_min)**2, v_min, v_max)),
Leaf(idx, "med",
fl.TriangularSet(v_min, v_min + ((v_max - v_min) / 2), v_max)),
Leaf(idx, "hig",
fl.TriangularSet(v_min, v_max, v_max + (v_max - v_min)**2))
]
def fuzzify_mean(A):
# output for fuzzified values
R = np.zeros((A.shape[0], A.shape[1] * 3))
cmin, cmax, cmean = np.nanmin(A, 0), np.nanmax(A, 0), np.nanmean(A, 0)
left = np.array([cmin - (cmax - cmin), cmin, cmax]).T
middle = np.array([cmin, cmean, cmax]).T
right = np.array([cmin, cmax, cmax + (cmax - cmin)]).T
mus = []
for i in range(A.shape[1]):
f_l = fl.TriangularSet(*left[i])
f_m = fl.TriangularSet(*middle[i])
f_r = fl.TriangularSet(*right[i])
R[:,(i*3)] = f_l(A[:,i])
R[:,(i*3)+1] = f_m(A[:,i])
R[:,(i*3)+2] = f_r(A[:,i])
mus.extend([(i, f_l), (i, f_m), (i, f_r)])
return 3, R, mus
def fuzzify_p(A):
R = np.zeros((A.shape[0], A.shape[1] * p))
cmin, cmax = np.nanmin(A, 0), np.nanmax(A, 0)
psize = (cmax - cmin) / (p - 1)
mus = []
# iterate features
for i in range(A.shape[1]):
# iterate partitions
mu_i = []
offset = cmin[i]
for j in range(p):
f = fl.TriangularSet(offset - psize[i], offset, offset + psize[i])
R[:, (i * p) + j] = f(A[:, i])
mu_i.append(f)
offset += psize[i]
mus.append(mu_i)
return p, R, mus
def build_t_membership(chromosome, idx):
a, b, c = sorted(chromosome[idx:idx + 3])
return fl.TriangularSet(a, b, c)
def t_factory(**k):
return fl.TriangularSet(k["min"], k["mean"], k["max"])
def triangular_factory(*args):
return fl.TriangularSet(args[0], args[1], args[2])
