秦越人的博客

Practice Tests Truth

UCINET社交网络分析

发表于 更新于 分类于 ，
本文字数： 1.7k 阅读时长 ≈ 6 分钟

1.共现矩阵

灰度共生矩阵

  • 先将图像转换成灰度图片
  • 然后再用graycomatirx函数进行共生矩阵计算
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
# -*- coding: utf-8 -*-
import cv2
import numpy as np
np.set_printoptions(suppress=True)

def glcm(arr, d_x, d_y, gray_level=16):
    '''计算并返回归一化后的灰度共生矩阵'''
    max_gray = arr.max()
    height, width = arr.shape
    arr = arr.astype(np.float64)  # 将uint8类型转换为float64，以免数据失真
    arr = arr * (gray_level - 1) // max_gray  # 若灰度级数大于gray_level，则将图像的灰度级缩小至gray_level，减小灰度共生矩阵的大小。量化后灰度值范围：0 ~ gray_level - 1
    ret = np.zeros([gray_level, gray_level])
    for j in range(height -  abs(d_y)):
        for i in range(width - abs(d_x)):  # range(width - d_x)  #注释为源代码，经评论指出错误后修改
            rows = arr[j][i].astype(int)
            cols = arr[j + d_y][i + d_x].astype(int)
            ret[rows][cols] += 1
    if d_x >= d_y:
        ret = ret / float(height * (width - 1))  # 归一化, 水平方向或垂直方向
    else:
        ret = ret / float((height - 1) * (width - 1))  # 归一化, 45度或135度方向
    return ret

if __name__=='__main__':
    '''归一化时分母值根据角度theta变化，0度或90度时为height * (width - 1), 45度或135度时为(height - 1) * (width - 1)'''
    fp = r'/home/jovyan/work/000.png'
    img = cv2.imread(fp)
    img_gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)  # 转换为灰度图像，uint8
    #glcm_0 = glcm(img_gray, 1, 0)  # 水平方向
    glcm_1 = glcm(img_gray, 0, 1)  # 垂直方向
    # glcm_2 = glcm(img_gray, 1, 1)  # 45度方向
    # glcm_3 = glcm(img_gray, -1, 1)  # 135度方向
    print(glcm_1)

    np.save('/home/jovyan/work/glcm_1.npy', glcm_1)

我们将npy.文件格式转换为csv，再转化为Excel文件

1
2
3
4
5
6
7
8
9
10
11
loaded_data = np.load('/home/jovyan/work/glcm_1.npy') 

import pandas as pd  
  
df = pd.DataFrame(loaded_data)

# 保存为CSV文件  
df.to_csv('/home/jovyan/work/output.csv', index=False)  
  
# 保存为Excel文件（需要安装openpyxl或xlsxwriter库）  
df.to_excel('output.xlsx', index=False)

输出的表格如下图：

image-20240323185009666

2 Ucinet分析

2.1 输入矩阵中

菜单栏第三行第二个

image-20240323185402089

image-20240323185201093

随后保存

image-20240323185323487

2.2 二值化

transform->dichotomize

    0 1 2 3 4 5 6 7 8 9 1 1 1 1 1 1
    - - - - - - - - - - - - - - - -
 0  1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0
 1  1 1 1 1 1 1 1 1 0 1 1 1 0 0 0 0
 2  1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0
 3  1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0
 4  1 1 1 1 1 1 1 1 1 1 1 1 0 1 0 0
 5  0 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0
 6  0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0
 7  0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0
 8  0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 0
 9  0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 0
10  0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
11  0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1
12  0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1
13  0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1
14  0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1
15  0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1

2.3 生成网络

image-20240323185515725

将刚才保存的共现矩阵导入

image-20240323185610750

image-20240323185751157

2.4 中心性分析

Analysis->centralitymeasures

img

image-20240323193324338

2.5 节点中心度

image-20240323185849724

FREEMAN’S DEGREE CENTRALITY MEASURES:

Diagonal valid? NO
Model: SYMMETRIC
Input dataset: cocomarix (C:\Users\lenovo\Desktop\论文\UCINET\cocomarix)

​ 1 2 3
​ Degree NrmDegree Share

8 7 0.017 13.909 0.130
7 6 0.017 13.291 0.124
9 8 0.014 10.977 0.103
15 14 0.011 8.892 0.083
10 9 0.010 8.241 0.077
14 13 0.010 8.011 0.075
6 5 0.009 7.025 0.066
11 10 0.007 5.887 0.055
13 12 0.007 5.866 0.055
4 3 0.007 5.270 0.049
12 11 0.006 4.755 0.045
16 15 0.005 3.821 0.036
3 2 0.005 3.633 0.034
5 4 0.004 3.246 0.030
2 1 0.003 2.043 0.019
1 0 0.002 1.904 0.018

DESCRIPTIVE STATISTICS

​ 1 2 3
​ Degree NrmDegree Share

1 Mean 0.008 6.673 0.063
2 Std Dev 0.004 3.572 0.033
3 Sum 0.133 106.771 1.000
4 Variance 0.000 12.763 0.001
5 SSQ 0.001 916.710 0.080
6 MCSSQ 0.000 204.204 0.018
7 Euc Norm 0.038 30.277 0.284
8 Minimum 0.002 1.904 0.018
9 Maximum 0.017 13.909 0.130

Network Centralization = 8.27%
Heterogeneity = 8.04%. Normalized = 1.91%

Actor-by-centrality matrix saved as dataset FreemanDegree

Running time: 00:00:01
Output generated: 23 3月 24 18:59:01
Copyright (c) 2002-8 Analytic Technologies

NrmDegree 标准中心度

Mean 均值

Std Dev 标准差

Variance 方差

SSQ 平方差和

MCSSQ 平均值平方差和

Euc Norm 欧几里得范数

2.6 接近中心度

network->centrality->closeness

CLOSENESS CENTRALITY

Input dataset: cocomarix (C:\Users\lenovo\Desktop\论文\UCINET\cocomarix)
Method: Geodesic paths only (Freeman Closeness)
Output dataset: Closeness (C:\Users\lenovo\Desktop\论文\UCINET\Closeness)

Note: Data not symmetric, therefore separate in-closeness & out-closeness computed.

Closeness Centrality Measures

               1            2            3            4
       inFarness   outFarness  inCloseness outCloseness
    ------------ ------------ ------------ ------------

11 10 16.000 16.000 93.750 93.750
10 9 16.000 18.000 93.750 83.333
7 6 16.000 17.000 93.750 88.235
12 11 17.000 18.000 88.235 83.333
6 5 17.000 19.000 88.235 78.947
8 7 17.000 17.000 88.235 88.235
9 8 18.000 18.000 83.333 83.333
5 4 18.000 18.000 83.333 83.333
4 3 18.000 20.000 83.333 75.000
13 12 19.000 18.000 78.947 83.333
3 2 20.000 18.000 75.000 83.333
14 13 20.000 21.000 75.000 71.429
15 14 21.000 22.000 71.429 68.182
2 1 22.000 20.000 68.182 75.000
16 15 26.000 25.000 57.692 60.000
1 0 29.000 25.000 51.724 60.000

Statistics

                     1            2            3            4
             inFarness   outFarness  inCloseness outCloseness
          ------------ ------------ ------------ ------------

1 Mean 19.375 19.375 79.621 78.674
2 Std Dev 3.569 2.595 12.153 9.445
3 Sum 310.000 310.000 1273.930 1258.778
4 Variance 12.734 6.734 147.693 89.217
5 SSQ 6210.000 6114.000 103794.195 100460.148
6 MCSSQ 203.750 107.750 2363.083 1427.469
7 Euc Norm 78.804 78.192 322.171 316.954
8 Minimum 16.000 16.000 51.724 60.000
9 Maximum 29.000 25.000 93.750 93.750

Network in-Centralization = 31.22%
Network out-Centralization = 33.31%

Output actor-by-centrality measure matrix saved as dataset Closeness (C:\Users\lenovo\Desktop\论文\UCINET\Closeness)

Running time: 00:00:01
Output generated: 23 3月 24 19:37:04
Copyright (c) 1999-2008 Analytic Technologies

2.7 中间中心度

network->centrality->freeman betweenness->node betweenness

FREEMAN BETWEENNESS CENTRALITY

Input dataset: cocomarix (C:\Users\lenovo\Desktop\论文\UCINET\cocomarix)

Important note: this routine binarizes but does NOT symmetrize.

Un-normalized centralization: 115.007

               1            2
     Betweenness nBetweenness
    ------------ ------------

11 10 11.563 5.506
7 6 7.874 3.749
5 4 6.907 3.289
12 11 6.215 2.959
10 9 5.533 2.635
4 3 5.395 2.569
8 7 4.846 2.308
3 2 4.621 2.201
13 12 4.389 2.090
6 5 3.718 1.771
9 8 2.939 1.400
14 13 2.330 1.109
2 1 2.310 1.100
15 14 1.359 0.647
1 0 0.000 0.000
16 15 0.000 0.000

DESCRIPTIVE STATISTICS FOR EACH MEASURE

                     1            2
           Betweenness nBetweenness
          ------------ ------------

1 Mean 4.375 2.083
2 Std Dev 2.901 1.382
3 Sum 70.000 33.333
4 Variance 8.418 1.909
5 SSQ 440.937 99.986
6 MCSSQ 134.687 30.541
7 Euc Norm 20.999 9.999
8 Minimum 0.000 0.000
9 Maximum 11.563 5.506

Network Centralization Index = 3.65%

Output actor-by-centrality measure matrix saved as dataset FreemanBetweenness

Running time: 00:00:01
Output generated: 23 3月 24 19:38:10
Copyright (c) 1999-2008 Analytic Technologies

2.8 凝聚子群分析

network->roles& positions ->structural ->concor

image-20240323193930316

3 结论

社交网络分析在灰度共生矩阵上的应用，反应了像素点之间的关联。从而可以得出图片中的痕迹是否明显。

4 与自己方向相结合

我的研究方向是视频深度伪造检测

本次提供的灰度共生数据，来源于以下图片

image-20240325082027764

倘若，再加以处理，例如：将真的图片中的人脸提取出来，再进行灰度共生矩阵处理，对其进行中心性分析等一系列社交网络分析。然后将所得的分析数据输入神经网络模型中，并重复输入不同的人脸图像，让模型进行学习。再将假图片的分析数据给训练后的模型进行评估。

-------------已经到底啦！-------------