NetViewer/scheduler/algo1Folder/AUC.py

import numpy as np
import random
import time

def Calculation_AUC(MatrixAdjacency_Train, MatrixAdjacency_Test, Matrix_similarity, MaxNodeNum):
    AUC_TimeStart = time.clock()
    print
    '    Calculation AUC......'
    AUCnum = 672400

    Matrix_similarity = np.triu(Matrix_similarity - Matrix_similarity * MatrixAdjacency_Train)
    Matrix_NoExist = np.ones(MaxNodeNum) - MatrixAdjacency_Train - MatrixAdjacency_Test - np.eye(MaxNodeNum)

    Test = np.triu(MatrixAdjacency_Test)
    NoExist = np.triu(Matrix_NoExist)

    #     Test_num =len(np.argwhere(Test == 1))
    #     NoExist_num = len(np.argwhere(NoExist == 1))
    # # #   Test_num = np.nonzero(Test)[0].shape[0]
    # # #   NoExist_num = np.nonzero(NoExist)[0].shape[0]

    Test_num = len(np.argwhere(Test == 1))
    NoExist_num = len(np.argwhere(NoExist == 1))
    #     print '    Test_num：%d'%Test_num
    #     print '    NoExist_num：%d'%NoExist_num

    Test_rd = [int(x) for index, x in enumerate((Test_num * np.random.rand(1, AUCnum))[0])]
    NoExist_rd = [int(x) for index, x in enumerate((NoExist_num * np.random.rand(1, AUCnum))[0])]
    #     print '    Test_rd：'+str(Test_rd)
    #     print '    Test_rd长度：'+str(len(Test_rd))
    #     print '    Test_rd最大值：'+str(max(Test_rd))
    #     print '    NoExist_rd：'+str(NoExist_rd)
    #     print '    NoExist_rd长度：'+str(len(NoExist_rd))
    TestPre = Matrix_similarity * Test
    NoExistPre = Matrix_similarity * NoExist

    TestIndex = np.argwhere(Test == 1)
    Test_Data = np.array([TestPre[x[0], x[1]] for index, x in enumerate(TestIndex)]).T
    NoExistIndex = np.argwhere(NoExist == 1)
    NoExist_Data = np.array([NoExistPre[x[0], x[1]] for index, x in enumerate(NoExistIndex)]).T
    #     print Test_Data
    #     print Test_Data.shape
    #     print NoExist_Data
    #     print NoExist_Data.shape

    Test_rd = np.array([Test_Data[x] for index, x in enumerate(Test_rd)])
    NoExist_rd = np.array([NoExist_Data[x] for index, x in enumerate(NoExist_rd)])
    #     print Test_rd
    #     print Test_rd.shape
    #     print NoExist_rd
    #     print NoExist_rd.shape

    #     aucArray = Test_rd - NoExist_rd
    #     n1 = len(np.argwhere(aucArray > 0))
    #     n2 = len(np.argwhere(aucArray == 0))
    n1, n2 = 0, 0
    for num in range(AUCnum):
        if Test_rd[num] > NoExist_rd[num]:
            n1 += 1
        elif Test_rd[num] == NoExist_rd[num]:
            n2 += 0.5
        else:
            n1 += 0
    auc = float(n1 + n2) / AUCnum
    print('    AUC指标为：%f' % auc)
    AUC_TimeEnd = time.clock()
    print('    AUCTime：%f s' % (AUC_TimeEnd - AUC_TimeStart))
    return auc


# 随机选择元素进行比较，并计算得分
def calculate_score(N, train, text, score, n):
    total_score = 0.0
    for i in range(n):
        while True:
            random_row = random.randint(0, N-1)
            random_col = random.randint(0, N-1)
            if train[random_row][random_col] == 0:
                rand_index_train = [random_row, random_col]
                break
        while True:
            random_row = random.randint(0, N-1)
            random_col = random.randint(0, N-1)
            if text[random_row][random_col] == 1:
                rand_index_text = [random_row, random_col]
                break

        # 计算得分
        ##print("text_score:", score[rand_index_text[0]][rand_index_text[1]])
        ##print("train_score:", score[rand_index_train[0]][rand_index_train[1]])
        if score[rand_index_text[0]][rand_index_text[1]] > score[rand_index_train[0]][rand_index_train[1]]:
            total_score += 1.0
        elif score[rand_index_text[0]][rand_index_text[1]] == score[rand_index_train[0]][rand_index_train[1]]:
            total_score += 0.5

    return total_score