一.实验目的
1.掌握贝叶斯分类器的基本原理与实现方法。
二.实验内容
1.编写程序实现贝叶斯分类器,完成基于西瓜书数据集3.0的贝叶斯分类器设计,并对P151所给定的测试1样本进行分类。
01 加载数据
%% 加载数据
D = importdata('data.txt');
n = 17; % 数据量
m = 10; % 属性
train = cell(n,m);
test = cell(1,m);
S = regexp(D,',','split');
for i=1:17
train(i,:) = S{i,1};
end
test(1,:) = S{n+1,1};
disp('加载训练数据:');
disp(train);
disp('加载测试数据:');
disp(test);
02 对于离散属性,求条件概率;对于连续属性,可以考虑概率密度函数。分别将测试集中每个属性在好瓜以及坏瓜中占有的比例进行计算并存储在元胞数组中。
%%
data = cell(3, m);
data(1,:) = test(1,:);
data{2,1} = '是';
data{3,1} = '否';
for i=2:3
for j=2:m
data{i,j} = 0;
end
end
%disp(an);
for i=2:7
for j=1:8
if train{j,i}==data{1,i}
data{2,i}=data{2,i}+1;
end
end
data{2,i}=data{2,i}./8;
for j=9:17
if train{j,i}==data{1,i}
data{3,i}=data{3,i}+1;
end
end
data{3,i}=data{3,i}./9;
end
data{2,10} = 8/17;
data{3,10} = 9/17;
% 计算密度和含糖率的概率密度
density1 = str2num(char(train(1:8,8)));
den_m1 = mean(density1);
den_s1 = std(density1);
data{2,8} = exp(-1*(str2num(char(test(1,8)))-den_m1)*(str2num(char(test(1,8)))-den_m1)/(2*den_s1*den_s1))/(sqrt(2*pi)*den_s1);
density2 = str2num(char(train(9:17,8)));
den_m2 = mean(density2);
den_s2 = std(density2);
data{3,8} = exp(-1*(str2num(char(test(1,8)))-den_m2)*(str2num(char(test(1,8)))-den_m2)/(2*den_s2*den_s2))/(sqrt(2*pi)*den_s2);
suger1 = str2num(char(train(1:8,9)));
sug_m1 = mean(suger1);
sug_s1 = std(suger1);
data{2,9} = exp(-1*(str2num(char(test(1,9)))-sug_m1)*(str2num(char(test(1,9)))-sug_m1)/(2*sug_s1*sug_s1))/(sqrt(2*pi)*sug_s1);
suger2 = str2num(char(train(9:17,9)));
sug_m2 = mean(suger2);
sug_s2 = std(suger2);
data{3,9} = exp(-1*(str2num(char(test(1,9)))-sug_m2)*(str2num(char(test(1,9)))-sug_m2)/(2*sug_s2*sug_s2))/(sqrt(2*pi)*sug_s2);
disp('在好瓜坏瓜的条件下的概率:');
disp(data);
03 相乘得结果
% 好瓜坏瓜的概率
res_is = 1.0;
for i=2:10
res_is = res_is .* cell2mat(data(2,i));
end
disp('好瓜的概率:');
disp(res_is);
res_not = 1.0;
for i=2:10
res_not = res_not .* cell2mat(data(3,i));
end
disp('坏瓜的概率:');
disp(res_not);