def calculate_and_plot_feature_selection_score(input, selection, name):
outputs = helper.input_normalization(input, a, count)
fig = plt.figure()
labels = ["brak normalizacji", "normalizacja"]
i = 0
for item in outputs: #normalizacja i brak
final = []
final_2 = []
cv_scores = []
for how_many_attrs in features: #range(1,32):#pierwsze 30 najlepszych cech
final.append(selection(item, how_many_attrs, cv_scores))
for item in final:
final_2.append(
helper.cross_validation(item,
[KNeighborsClassifier(n_neighbors=5)]))
# print(final_2)
all_scores.append(final_2)
plt.plot(features, final_2, label=labels[i])
i += 1
plt.title(name + " dla algorytmu 5-najbliszych sasiadow")
plt.legend()
ax = fig.gca()
plt.xlabel("Liczba Cech")
plt.ylabel("Dokladnosc")
ax.set_xticks(features)
ax.set_yticks(np.arange(0, 1., 0.1))
plt.grid()
def k_choice(input):
fig = 1
string_index = 0
k_chosen = []
normal_string = []
scores = []
for p in [1, 2]:
cv_scores = []
for i in range(1, 10):
scores_mean = (helper.cross_validation(
input,
KNeighborsClassifier(n_neighbors=i, p=p,
metric='minkowski')), i)
cv_scores.append(scores_mean)
scores.append(scores_mean)
Range = list(range(1, 10))
cv_scores = [x[0] for x in cv_scores]
plt.subplot(4, 2, fig)
if p == 1:
plt.title("Manhattan distance measurement " +
normal_string[string_index])
if p == 2:
plt.title("Euclidean distance measurement " +
normal_string[string_index])
plt.plot(Range, cv_scores)
plt.xlabel('Number of Neighbors K')
plt.ylabel('Learning level')
fig += 1
string_index += 1
final_scores = sorted(scores, key=lambda x: x[0], reverse=True)
k_chosen.append(final_scores)
for i in range(4):
print(k_chosen[0][i][1])
plt.show()
import sys
sys.path.append("../")
from helper.utils import pd, Print, random_seed_cpu
from helper.cross_validation import *
from config import *
from tqdm import tqdm
random_seed_cpu(RANDOM_STATE)
target = pd.read_pickle(join(proc_data_path, "target.p"))
target = target[~target.label.isna()]
CV = cross_validation(train_df=target,
_id_="fid",
target_name="label",
kfold_type="skfold",
output_dir=proc_data_path,
split_ratio=0.1,
nfolds=5,
random_state=RANDOM_STATE,
shuffle=True,
stratify=False,
tag="_fold_5",
group_name="fid")
target = CV.split()