def normalize_data_2d():
input_values, contaminated = ia_n2020.ioValues()
predict_values = ia_n2020.newData()
norm = StandardScaler()
input_values_norm = norm.fit_transform(input_values)
predict_values_norm = norm.fit_transform(predict_values)
pca = PCA(n_components=2)
input_values_2d = pca.fit_transform(input_values_norm)
predict_values_2d = pca.fit_transform(predict_values_norm)
reg1 = ''
reg2 = ''
for i in range(len(input_values_2d)):
if contaminated[i] == 0:
reg1 = plt.scatter(input_values_2d[i][0],
input_values_2d[i][1],
marker='x',
color='g')
elif contaminated[i] == 1:
reg2 = plt.scatter(input_values_2d[i][0],
input_values_2d[i][1],
marker='o',
color='b')
plt.xlabel('PC1')
plt.ylabel('PC2')
plt.grid(True)
plt.legend((reg1, reg2), ('Não contaminado', 'Contaminado'))
plt.savefig('graphs/pca_graph.png')
plt.close()
return input_values_2d, predict_values_2d
def normalize_data_3d():
input_values, contaminated = ia_n2020.ioValues()
predict_values = ia_n2020.newData()
norm = StandardScaler()
input_values_norm = norm.fit_transform(input_values)
predict_values_norm = norm.fit_transform(predict_values)
pca = PCA(n_components=3)
input_values_3d = pca.fit_transform(input_values_norm)
predict_values_3d = pca.fit_transform(predict_values_norm)
return input_values_3d, predict_values_3d
from sklearn import svm
import pandas as pd
import ia_n2020
import pca_graph
input_values, contaminated = ia_n2020.ioValues()
predict_values = ia_n2020.newData()
input_values_2d, predict_values_2d = pca_graph.normalize_data_2d()
input_values_3d, predict_values_3d = pca_graph.normalize_data_3d()
# svm classifier with 4 input data
# sv_clf = svm.SVC(kernel='linear', C=0.5)
sv_clf = svm.SVC(kernel='rbf', gamma=1, C=0.5)
# sv_clf = svm.SVC(kernel='poly', degree=3, C=0.5)
sv_clf.fit(input_values, contaminated)
sv_score = sv_clf.score(input_values, contaminated)
print(sv_score)
score_sheet = pd.read_csv('output_data/score.csv')
score_sheet['sv_score'] = sv_score
score_sheet.to_csv('output_data/score.csv', index=False)
sv_predict = sv_clf.predict(predict_values)
print(sv_predict)
spreadsheet = pd.read_csv('output_data/predict_data.csv')
spreadsheet['svm_predict'] = sv_predict
spreadsheet.to_csv('output_data/predict_data.csv', index=False)
# svm classifier with 2D data
sv_clf_2d = svm.SVC(kernel='linear', C=0.5)
# sv_clf_2d = svm.SVC(kernel='rbf', gamma=1, C=0.5)