Prediction of red wine quality by decision tree

% Import excel data
filename="D:\Matlab\bin\ Homework \B topic The attachment 1.xlsx";
Title=[" Non volatile acid content "," Volatile acid content "," Citric acid "," Sugar content "," Chloride "," Free sulfur dioxide "," Total sulfur dioxide "," density ","PH value "," Sulfates "," alcohol "];
xlswrite(filename,Title);
[data,top,all]=xlsread(filename);
[data1,top1,all1]=xlsread(filename,' Data set to be predicted ');
%------------------------------------------------------------------------------------------------------------------------------------
% Data sets are preprocessed
%rmmissing Delete missing value processing
list=rmmissing(data); 
list1=rmmissing(data1); 
% normalization z-score Standardized treatment
R = zscore(list);
%-------------------------------------------------------------------------------------------------------------------------------------
% Randomly generated training set / Test set
P=randperm(3867);
Train =list(P(1:3000),:);
rest=list(P(3001:end),:);
% Training data
P_Train=Train(:,1:11);
T_Train=Train(:,12);
% Test data
P_rest=rest(:,1:11);
T_rest=rest(:,12);
% Create a decision tree classifier
ctree=ClassificationTree.fit(P_Train,T_Train);
% View decision tree view
view(ctree);
view(ctree,'mode','graph');
% The simulation test
T_sim=predict(ctree,P_rest);
%V. Result analysis
count_3= length(find(T_Train == 3));rate_3=count_3/3700;total_3=length(find(list(:,12)==3));
count_4= length(find(T_Train == 4));rate_4=count_4/3700;total_4=length(find(list(:,12)==4));
count_5= length(find(T_Train == 5));rate_5=count_5/3700;total_5=length(find(list(:,12)==5));
count_6= length(find(T_Train == 6));rate_6=count_6/3700;total_6=length(find(list(:,12)==6));
count_7= length(find(T_Train == 7));rate_7=count_7/3700;total_7=length(find(list(:,12)==7));
count_8= length(find(T_Train == 8));rate_8=count_8/3700;total_8=length(find(list(:,12)==8));
count_9= length(find(T_Train == 9));rate_9=count_9/3700;total_9=length(find(list(:,12)==9));
number_3= length(find(T_rest == 3));number_B3_sim= length(find(T_sim ==3&T_rest == 3));
number_4= length(find(T_rest == 4));number_B4_sim= length(find(T_sim ==4&T_rest == 4));
number_5= length(find(T_rest == 5));number_B5_sim= length(find(T_sim ==5&T_rest == 5));
number_6= length(find(T_rest == 6));number_B6_sim= length(find(T_sim ==6&T_rest == 6));
number_7= length(find(T_rest == 7));number_B7_sim= length(find(T_sim ==7&T_rest == 7));
number_8= length(find(T_rest == 8));number_B8_sim= length(find(T_sim ==8&T_rest == 8));
number_9= length(find(T_rest == 9));number_B9_sim= length(find(T_sim ==9&T_rest == 9));
fprintf(' Total red wine measurement ：%d\n',3867);
fprintf(' The quality is 3：%d\n',total_3);
fprintf(' The quality is 4：%d\n',total_4);
fprintf(' The quality is 5：%d\n',total_5);
fprintf(' The quality is 6：%d\n',total_6);
fprintf(' The quality is 7：%d\n',total_7);
fprintf(' The quality is 8：%d\n',total_8);
fprintf(' The quality is 9：%d\n',total_9);
fprintf(' The total number of red wine measurements in the training set ：%d\n',3700);
fprintf(' The quality is 3：%d\n',count_3);
fprintf(' The quality is 4：%d\n',count_4);
fprintf(' The quality is 5：%d\n',count_5);
fprintf(' The quality is 6：%d\n',count_6);
fprintf(' The quality is 7：%d\n',count_7);
fprintf(' The quality is 8：%d\n',count_8);
fprintf(' The quality is 9：%d\n',count_9);
fprintf(' The total number of red wine measurements in the test set ：%d\n',167);
fprintf(' The quality is 3：%d\n',number_3);
fprintf(' The quality is 4：%d\n',number_4);
fprintf(' The quality is 5：%d\n',number_5);
fprintf(' The quality is 6：%d\n',number_6);
fprintf(' The quality is 7：%d\n',number_7);
fprintf(' The quality is 8：%d\n',number_8);
fprintf(' The quality is 9：%d\n',number_9);
fprintf(' The number of correct quality predictions :%d\n',number_B3_sim+number_B4_sim+number_B5_sim+number_B6_sim+number_B7_sim+number_B8_sim+number_B9_sim);
fprintf(' Wrong number %d\n',number_3+number_4+number_5+number_6+number_7+number_8+number_9-(number_B3_sim+number_B4_sim+number_B5_sim+number_B6_sim+number_B7_sim+number_B8_sim+number_B9_sim));
fprintf(' Accuracy rate p:%f%%\n',(number_B3_sim+number_B4_sim+number_B5_sim+number_B6_sim+number_B7_sim+number_B8_sim+number_B9_sim)/(number_3+number_4+number_5+number_6+number_7+number_8+number_9)*100);  
%. The influence of the minimum number of samples contained in leaf nodes on the performance of decision tree
  leafs= logspace(1,2,10);
   N= numel(leafs);
   err= zeros(N,1);
  for n= 1:N
    t= ClassificationTree.fit(P_Train,T_Train,'crossval','on','minleaf',leafs(n));
    err(n)= kfoldLoss(t);
  end
   plot(leafs,err);
   xlabel(' The minimum number of samples contained in leaf nodes ');
   ylabel(' Cross validation error ');
   title(' The influence of the minimum number of samples contained in leaf nodes on the performance of decision tree ');  
% Set up minleaf by 10, Generate optimization decision tree
OptimalTree=ClassificationTree.fit(P_Train,T_Train,'minleaf',10);
view(OptimalTree,'mode','graph'); 
%1. The resampling error and cross validation error of the optimized decision tree are calculated
resubOpt= resubLoss(OptimalTree);
lossOpt= kfoldLoss(crossval(OptimalTree));
%2. The resampling error and cross validation error of the decision tree before optimization are calculated
resubDefault= resubLoss(ctree);
lossDefault= kfoldLoss(crossval(ctree));
%%. prune
[~,~,~,bestlevel]=cvLoss(ctree,'subtrees','all','treesize','min');
cptree= prune(ctree,'Level',bestlevel);
view(cptree,'mode','graph'); 
%1. The resampling error and cross validation error of the decision tree after pruning are calculated
resubPrune= resubLoss(cptree);
lossPrune= kfoldLoss(crossval(cptree));
%------------------------------------------------------------------------------------------------------------------------------------
% surface 2 data
% Randomly generated training set / Test set
P1=randperm(2000);
Train1 =list(P1(1:1000),:);
rest1=list1(1:1000,:);
% Training data
P_Train1=Train1(:,1:11);
T_Train1=Train1(:,12);
% Test data
P_rest1=rest1(:,1:11);
% Create a decision tree classifier
ctree1=ClassificationTree.fit(P_Train1,T_Train1);
% View decision tree view
view(ctree1);
view(ctree1,'mode','graph');
% The simulation test
T_sim1=predict(ctree1,P_rest1);
%------------------------------------------------------------------------------------------------------------------------------------
% Forward processing steps
[a,b]=size(list);
disp([' share ' num2str(a) ' Two evaluation objects , ' num2str(b-1) ' Evaluation indicators ']);
position = input(' Please enter the column of the indicator that needs forward processing , for example [2,4,5]：');
disp(' Please enter the column indicator type to be processed (1: Very small ,2： The middle type , 3： Interval type )')
type = input(' For example 2 Columns are very small , The first 4 Columns are interval type , The first 5 Columns are intermediate , Input [1,3,2]: ');
 % Be careful ,Position and Type Are two row vectors of the same dimension
 X=list;
    for i = 1 : size(position,2)  % These columns need to be treated separately , So we need to know the total number of times to process , That is, the number of cycles
        X(:,position(i)) = Positivization(X(:,position(i)),type(i),position(i));
    end
    disp(' The forward matrix X =  ')
    disp(X)
weigh=[0.15;0.15;0.1;0.2;0.05;0.05;0.05;0.05;0.01;0.01;0.18];% The weight
R(:,12)=[];
r=R';
D_P = sum(((r(:,1:a)-repmat(max(r(:,1:a)),1,1)) .^2 ) .* repmat(weigh,1,a) ,2) .^ 0.5;   % D+ The distance vector from the maximum
D_N = sum(((r(:,1:a)-repmat(min(r(:,1:a)),1,1)) .^2 ) .* repmat(weigh,1,a) ,2) .^ 0.5;   % D- The distance vector from the minimum
S = D_N ./ (D_P+D_N);    % Non normalized score
disp(' The final score is ：');
stand_S = S / sum(S);
[sorted_S,index] = sort(stand_S ,'descend');
disp(index);

Reference function Positivization.m

% function [ Output variables ] = The name of the function ( The input variable ）  
function [posit_x] = Positivization(x,type,i)
% There are three input variables ：
% x： The original column vector corresponding to the indicator that needs forward processing
% type： Types of indicators （1： Very small , 2： The middle type , 3： Interval type ）
% i: Which column in the original matrix is being processed
% Output variables posit_x Express ： The forward column vector
    if type == 1  % Very small
        disp([' The first ' num2str(i) ' Columns are very small , Is moving forward '] )
        posit_x = max(x) - x;
        disp([' The first ' num2str(i) ' Column miniaturization forward processing is complete '] )
        disp('-------------------- Demarcation line --------------------')
    elseif type == 2  % The middle type
        disp([' The first ' num2str(i) ' Columns are intermediate '] )
        best = input(' Please enter the best value ： ');
        M = max(abs(best-x));
        posit_x = 1-abs(best-x)/M;
        disp([' The first ' num2str(i) ' The forward processing of column intermediate type is completed '] )
        disp('-------------------- Demarcation line --------------------')
    elseif type == 3  % Interval type
        disp([' The first ' num2str(i) ' Columns are interval type '] )
        a = input(' Please enter the lower bound of the interval ： ');
        b = input(' Please enter the upper bound of the interval ： '); 
        r_data = size(x,1);
        M = max([a-min(x),max(x)-b]);
        posit_x = zeros(r_data,1);
        for i = 1:r_data
            if x(i)<a
                posit_x(i) = 1-(a-x(i))/M;
            elseif x(i)>b
                posit_x(i) = 1-(x(i)-b)/M;
            else
                posit_x(i) = 1;
            end
        end
        disp([' The first ' num2str(i) ' Column interval type forward processing completed '] )
        disp('-------------------- Demarcation line --------------------')
    else
        disp(' There are no indicators of this type , Please check Type Is there anything in the vector other than 1、2、3 Other values ')
    end
end