【数学建模】基于matlab模糊二元决策树【含Matlab源码 038期】

一、获取代码方式

获取代码方式1：
完整代码已上传我的资源：【数据建模】基于matlab模糊二元决策树【含Matlab源码 038期】

获取代码方式2：
通过订阅紫极神光博客付费专栏，凭支付凭证，私信博主，可获得此代码。

备注：
订阅紫极神光博客付费专栏，可免费获得1份代码（有效期为订阅日起，三天内有效）；

二、简介

模糊模式树（FPT） 是一种机器学习的分类方法，类似于二元决策树（binary descision diagram,BDD）。其内部节点使用广义（模糊）逻辑和算数运算符标记，叶节点与给定输入属性集合上的（一元）模糊谓词相关联
模糊模式树（FPT）
在预测性能方面与其他机器学习竞争的方法相比是有竞争力的，由于物联网节点处理资源有限，他倾向于生成在解释方面非常有吸引力的紧凑模型。
其缺点也很明显:其计算复杂度高，运行时间可能变得令人无法接受。

二分决策树（BDD）
BDD是用来表达布尔函数的一种数据结构。是一种基于shannon分解的有向无环图。
设f为一组布尔变量集合的布尔函数（boolean function），且x为集合X中的变量，则shannon分解及其if-then-else(ite)形式为

式中x和 x-bar 分别代表事件发生，事件不发生；fx=1（即F1）和fx=0（即F2）
分别代表了事件发生与事件不发生的布尔函数
二叉树表示为以下：

布尔函数（boolean function）

布尔函数基于对布尔输入，通过某种逻辑计算，输出布尔值。
F（b1,b2,b3,…,bn）
带有n个两元素布尔代数{0,1}的布尔变量bi，F的值也是在{0,1}中
一般定义域上值在{0,1}中的函数也被称为布尔值函数，所以布尔函数是特殊情况

三、部分代码

%%----- Main Program that should be run inorder to execute the code -----%%
 
% load the data file
load('earfeature15window700image.mat');
 
 
% Change following variable in case of number of user changes
noOfUsers = 100;
 
% Change following two variable in case you need to change ration of test
% and train case or when the number of data value for same user changes(right now its 7)
trainCasePerUser = 4;
testCasePerUser = 3;
 
if(testCasePerUser < 2)
    disp('testCasePerUser should be greater then 1')
    
else  
totalCasePerUser = trainCasePerUser + testCasePerUser;
% 3D matrix where 1st dimention is related the user, 2nd dimention related
% to datavalues per user and 3rd dimention are the feature values for datapoints 
test = zeros([noOfUsers testCasePerUser 35]);
train = zeros([noOfUsers trainCasePerUser 35]);
%% this loop devides the data into testcases and trainingcase randomly
for i=1:noOfUsers
    
   % matrix p is the random permutation of interger 1 to total number of datavalues per user  
   p = randperm(totalCasePerUser,totalCasePerUser);
   
   % below two lines divide the data into tests and training data
   test(i,:,:) = T_S1(:,(i-1)*totalCasePerUser + p(1:testCasePerUser))';
   train(i,:,:) = T_S1(:,(i-1)*totalCasePerUser + p(testCasePerUser+1:totalCasePerUser))';
end
%% Following piece of code find creates the matching score matrix with 2 columns 
% 1st column has the value of matching score and 2nd column has value 1 if 
% its genuine and 0 if imposter
trainScore = zeros([noOfUsers*noOfUsers 2]);
testScore = zeros([noOfUsers*noOfUsers*(testCasePerUser-1) 2]);
% loop that creates the matrix of training sample.
for i=1:noOfUsers
    for j=1:noOfUsers
        score = zeros([trainCasePerUser 1]);
        for k = 1:trainCasePerUser
            score(k) = norm(squeeze((test(i,1,:) - train(j,k,:))),2);
        end
        
        % take the min of distance to get the matching score
        trainScore((i-1)*noOfUsers + j,1) = min(score);
        
        %  if the users are same then its genuine class else imposter
        if(i==j)
            trainScore((i-1)*noOfUsers + j,2) = 1;
        else
            trainScore((i-1)*noOfUsers + j,2) = 0;
        end        
    end
end
% loop that creates the matrix of testing sample
for z=0:testCasePerUser-2
    for i=1:noOfUsers
        for j=1:noOfUsers
            score = zeros([trainCasePerUser 1]);
            for k = 1:trainCasePerUser
                score(k) = norm(squeeze((test(i,z+2,:) - train(j,k,:))),2);
            end
            testScore(z*noOfUsers*noOfUsers + (i-1)*noOfUsers + j,1) = min(score);
            if(i==j)
                testScore(z*noOfUsers*noOfUsers + (i-1)*noOfUsers + j,2) = 1;
            else
                testScore(z*noOfUsers*noOfUsers + (i-1)*noOfUsers + j,2) = 0;
            end        
        end
    end
end
% find the geninue class datapoint and put them in Gtrain Matrix
Gtrain = trainScore(find(trainScore(:,2) == 1), :);
% find the imposter class datapoint and put them in Itrain Matrix
Itrain = trainScore(find(trainScore(:,2) == 0), :);
% Create the final training matrix with  score value in 1st column, membership 
% value in 2nd column and class in 3rd column for creating dicision tree.
% change the membership function 2nd  column in case you want to use
% different membership funtion
Mtrain = [Gtrain(:,1) GaussianMembership(Gtrain(:,1)) Gtrain(:,2); Itrain(:,1) TrapezoidalMembership(Itrain(:,1)) Itrain(:,2)];
%Mtrain = [Gtrain(:,1) TrapezoidalMembership(Gtrain(:,1)) Gtrain(:,2); Itrain(:,1) TrapezoidalMembership(Itrain(:,1)) Itrain(:,2)];
%% This part of code deals with growing the decision tree
% sort the training matrix according the matching score value.
[a b] = sort(Mtrain(:,1));
clear a;
data = Mtrain(b,:);
% grow the tree using sorted training matrix
tree = growTree( data );
%% This part of code deals with the predicting the class of testing score and
% calculation of FRR and FAR
testingSize = noOfUsers*noOfUsers*(testCasePerUser-1);
geniuneTestingSize = noOfUsers*(testCasePerUser-1);
trainingSize = noOfUsers*noOfUsers;
geniuneTrainingSize = noOfUsers;
pre = zeros([testingSize 2]);
% loop that predicts the values of all testing data
for i=1:testingSize
    pre(i,1) = predict(tree, testScore(i,:));
    pre(i,2) = testScore(i,2);
end
% calculation of FAR and FRR
FA = pre(find(pre(:,1) == 1 & pre(:,2) == 0),:); % False Acceptance
FR = pre(find(pre(:,1) == 0 & pre(:,2) == 1),:); % False Rejection
FAR = size(FA,1)/(testingSize-geniuneTestingSize); % False Acceptance Rate
FRR = size(FR,1)/geniuneTestingSize; % False Rejection Rate
%% histgram of normalized frequencies of matching score of Geniune and Imposter matching score
figure;
[m,yout] = hist(Itrain(:,1), 20);
m = m/(size(Itrain,1));
[n,xout] = hist(Gtrain(:,1), 20);
n = n/size(Gtrain,1);
bar(yout,m);
hold on;
bar(xout,n, 'r');
legend('Imposter Scores','Geniune Scores');
hold off;
end

  

 

  
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
  
36
  
37
  
38
  
39
  
40
  
41
  
42
  
43
  
44
  
45
  
46
  
47
  
48
  
49
  
50
  
51
  
52
  
53
  
54
  
55
  
56
  
57
  
58
  
59
  
60
  
61
  
62
  
63
  
64
  
65
  
66
  
67
  
68
  
69
  
70
  
71
  
72
  
73
  
74
  
75
  
76
  
77
  
78
  
79
  
80
  
81
  
82
  
83
  
84
  
85
  
86
  
87
  
88
  
89
  
90
  
91
  
92
  
93
  
94
  
95
  
96
  
97
  
98
  
99
  
100
  
101
  
102
  
103
  
104
  
105
  
106
  
107
  
108
  
109
  
110
  
111
  
112
  
113
  
114
  
115
  
116
  
117
  
118
  
119
  
120
 

四、备注

版本：2014a

文章来源: qq912100926.blog.csdn.net，作者：海神之光，版权归原作者所有，如需转载，请联系作者。

原文链接：qq912100926.blog.csdn.net/article/details/112153298