%% 清空环境变量

clear

clc

%% 训练集/测试集产生

% 导入数据

load water_data.mat

% 数据归一化

attributes = mapminmax(attributes);

% 训练集——35个样本

P_train = attributes(:,1:35);

T_train = classes(:,1:35);

% 测试集——4个样本

P_test = attributes(:,36:end);

T_test = classes(:,36:end);

%% 竞争神经网络创建、训练及仿真测试

% 创建网络

net = competlayer(4);

% 设置训练参数

net.trainParam.epochs = 500;

% 训练网络

net = train(net,P_train);

% 仿真测试

% 训练集

t_sim_compet_1 = sim(net,P_train);

T_sim_compet_1 = vec2ind(t_sim_compet_1);

% 测试集

t_sim_compet_2 = sim(net,P_test);

T_sim_compet_2 = vec2ind(t_sim_compet_2);

%% SOFM神经网络创建、训练及仿真测试

% 创建网络

net = selforgmap([4 4]);

% 设置训练参数

net.trainParam.epochs = 200;

% 训练网络

net = train(net,P_train);

% 仿真测试

% 训练集

t_sim_sofm_1 = sim(net,P_train);

T_sim_sofm_1 = vec2ind(t_sim_sofm_1);

% 测试集

t_sim_sofm_2 = sim(net,P_test);

T_sim_sofm_2 = vec2ind(t_sim_sofm_2);

%% 网络神经元分布情况

% 查看网络拓扑学结构

figure

plotsomtop(net)

% 查看临近神经元直接的距离情况

figure

plotsomnd(net)

% 查看每个神经元的分类情况

figure

plotsomhits(net,P_train)

%% 结果对比

% 竞争神经网络

result_compet_1 = [T_train' T_sim_compet_1']

result_compet_2 = [T_test' T_sim_compet_2']

% SOFM神经网络

result_sofm_1 = [T_train' T_sim_sofm_1']

result_sofm_2 = [T_test' T_sim_sofm_2']