def population_fitness(train_features, train_labels, test_features,
test_labels, population):
# 计算种群每个个体的自适应度值
population_fit = np.zeros(len(population))
for i in range(len(population)):
index = population[i]
train_feature_exp = train_features[:, index == 1]
test_features_exp = test_features[:, index == 1]
# predictY, accuracy = myKNN(k, trainX, trainY, testX, testY)
predictY, accuracy = myKNN(1, train_feature_exp, train_labels,
test_features_exp, test_labels)
population_fit[i] = accuracy
return population_fit
def kCrossValidation(allData,allLabel,trueLabel,countArr):
kf = KFold(n_splits=5, shuffle=True)
# m = len(countArr)
# print 'm is ', m
# temp = np.zeros((m, 1))
# for i in range(m):
# temp[i] = i
accuracy = np.array([0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0])
trueAccuracy = np.array([0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0])
for trainIndex, testIndex in kf.split(trueLabel):
trainData, trainLabel, trainCountArr, trainTrueLabel, testData, testLabel, testCountArr, testTrueLabel = getTrainTestData(allData, allLabel, trueLabel, countArr, trainIndex, testIndex)
a, b = knn.myKNN(trainData, trainLabel, testData, testLabel, testTrueLabel, testCountArr)
accuracy += a
trueAccuracy += b
accuracy = np.array(accuracy)/5
trueAccuracy = np.array(trueAccuracy)/5
print 'average Accuracy is ', accuracy
print 'average TrueAccuracy is', trueAccuracy
print 'allAverageTrueAccuracy is ', np.sum(trueAccuracy)/9
plt.plot(trueAccuracy)
plt.show()
import numpy as np
from myKNN import myKNN
trn_data = np.loadtxt('optdigits_train.txt', delimiter = ',')
tst_data = np.loadtxt('optdigits_test.txt', delimiter = ',')
training_feature_matrix = trn_data[:,:-1]
training_response = trn_data[:,-1]
test_feature_matrix = tst_data[:,:-1]
test_response = tst_data[:,-1]
k_values = [1,3,5,7]
error_rates_dict = {}
for x in k_values :
print("KNN for k value :", x)
predictions = myKNN(trn_data, tst_data, x)
error = 0
for j in range(len(predictions)) :
if(test_response[j] != predictions[j]) :
error += 1
print("Number of errors :", error)
error_rate = error / len(predictions)
print("Error rate for k :",x, " is = ", error_rate)
error_rates_dict[x] = error_rate
print("===============================================================================")
print(error_rates_dict)
print("Projection into dimension of size :", l)
print()
matrix_w, eigen_values = myLDA(trn_data, l)
train_lda = training_feature_matrix.dot(matrix_w)
test_lda = test_feature_matrix.dot(matrix_w)
train_lda_concat = np.hstack(
(train_lda, training_response.reshape(1500, 1)))
test_lda_concat = np.hstack((test_lda, test_response.reshape(297, 1)))
k_values = [1, 3, 5]
error_rates_dict = {}
for x in k_values:
print("KNN for k value :", x)
predictions = myKNN(train_lda_concat, test_lda_concat, x)
error = 0
for j in range(len(predictions)):
if (test_response[j] != predictions[j]):
error += 1
print("Number of errors :", error)
error_rate = error / len(predictions)
print("Error rate for k :", x, " is = ", error_rate)
error_rates_dict[x] = error_rate
print(
"==============================================================================="
)
print(error_rates_dict)
eigen_vec_count)
matrix_w, eigen_vals_pc = myPCA(trn_data, eigen_vec_count)
train_pca = training_feature_matrix.dot(matrix_w)
test_pca = test_feature_matrix.dot(matrix_w)
train_pca_concat = np.hstack((train_pca, training_response.reshape(500, 1)))
test_pca_concat = np.hstack((test_pca, test_response.reshape(124, 1)))
k_values = [1, 3, 5, 7]
error_rates_dict = {}
for x in k_values:
print("KNN for k value :", x)
predictions = myKNN(train_pca_concat, test_pca_concat, x)
error = 0
for j in range(len(predictions)):
if (test_response[j] != predictions[j]):
error += 1
print("Number of errors :", error)
error_rate = error / len(predictions)
print("Error rate for k :", x, " is = ", error_rate)
error_rates_dict[x] = error_rate
print(
"==============================================================================="
)
print(error_rates_dict)