def NEW_PTM_WKNN(data_set):
train, test = Pretreatment.partitioningTPMKNN(data_set)
normalized_train_matrix = Pretreatment.normalization(train)
normalized_test_matrix = Pretreatment.normalization(test)
klist = np.zeros((normalized_train_matrix.shape[0], 1))
predictions = np.zeros((normalized_test_matrix.shape[0], 1))
for unknowid in range(0,
len(normalized_train_matrix)): # 给训练集样本加入标签local k
klist[unknowid] = PTM_KNN.trainlocal_k(normalized_train_matrix,
unknowid)
new_normalized_train_matrix = np.column_stack(
(normalized_train_matrix, klist))
for unknowid_text in range(
0, len(normalized_test_matrix)): # 选取测试集样本最近三个点的最大local k作为最佳k
best_c = bestc(new_normalized_train_matrix, normalized_test_matrix,
unknowid_text, data_set) # 循环选取c值
new_k1 = PTM_KNN.testlocal_k(new_normalized_train_matrix,
normalized_test_matrix, unknowid_text,
best_c) #通过c值选取k值
predictions[unknowid_text] = WKNN.weighted_knn(
unknowid_text, normalized_train_matrix, int(new_k1),
normalized_test_matrix)
# 计算邻近k个值中分类权重最高的分类作为预测分类
TP, FP, FN, TN = analysis.TPFPFNTN(normalized_test_matrix, predictions)
return analysis.recall(TP,
FN), analysis.F_score(TP, FP, FN), analysis.G_mean(
TP, FN, TN, FP), TP, FN, TN, FP
def WKNN(data_set,best_k):
train, test = Pretreatment.partitioningTPMKNN(data_set)
normalized_train_matrix = Pretreatment.normalization(train)
normalized_test_matrix = Pretreatment.normalization(test)
predictions = np.zeros((normalized_test_matrix.shape[0], 1))
for unknowid_text in range(0, len(normalized_test_matrix)):
predictions[unknowid_text] = weighted_knn(unknowid_text, normalized_train_matrix, int(best_k), normalized_test_matrix)
TP, FP, FN, TN = analysis.TPFPFNTN(normalized_test_matrix, predictions)
return analysis.recall(TP, FN), analysis.F_score(TP, FP, FN), analysis.G_mean(TP, FN, TN, FP),TP, FN, TN, FP
def PTM_KNN(data_set):
train, test = Pretreatment.partitioningTPMKNN(data_set)
normalized_train_matrix = Pretreatment.normalization(train)
normalized_test_matrix = Pretreatment.normalization(test)
klist = np.zeros((normalized_train_matrix.shape[0], 1))
predictions = np.zeros((normalized_test_matrix.shape[0], 1))
for unknowid in range(0, len(normalized_train_matrix)):
klist[unknowid] = trainlocal_k(normalized_train_matrix, unknowid)
new_normalized_train_matrix = np.column_stack(
(normalized_train_matrix, klist))
for unknowid_text in range(0, len(normalized_test_matrix)):
new_k = testlocal_k(new_normalized_train_matrix,
normalized_test_matrix, unknowid_text, 3)
predictions[unknowid_text] = KNN.traditional_knn(
unknowid_text, normalized_train_matrix, int(new_k),
normalized_test_matrix)
TP, FP, FN, TN = analysis.TPFPFNTN(normalized_test_matrix, predictions)
return analysis.recall(TP,
FN), analysis.F_score(TP, FP, FN), analysis.G_mean(
TP, FN, TN, FP), TP, FN, TN, FP