随机森林.zip随机森林.zip

load imports-85 Y = X(:,1); X = X(:,2:end); isCategorical = [zeros(15,1);ones(size(X,2)-15,1)]; % Categorical variable flag % Y = PropVals.data(:,4); % X = Spectra.data; % isCategorical = zeros(size(X,2),1); % Categorical variable flag rng(1945,'twister') leaf = [5 10 20 50 100]; col = 'rbcmy'; figure for i=1:length(leaf) b = TreeBagger(50,X,Y,'Method','R','OOBPrediction','On',... 'CategoricalPredictors',find(isCategorical == 1),... 'MinLeafSize',leaf(i)); plot(oobError(b),col(i)) hold on end xlabel('Number of Grown Trees') ylabel('Mean Squared Error') legend({'5' '10' '20' '50' '100'},'Location','NorthEast') hold off b = TreeBagger(150,X,Y,'Method','R','OOBPredictorImportance','On',... 'CategoricalPredictors',find(isCategorical == 1),... 'MinLeafSize',5); figure plot(oobError(b)) xlabel('Number of Grown Trees') ylabel('Out-of-Bag Mean Squared Error') Yfit=predict(b,X); co=corr(Y,Yfit,'type','Pearson'); RMSE=sqrt(sum((Y-Yfit).^2)/size(Y,1)); % figure % bar(b.OOBPermutedPredictorDeltaError) % xlabel('Feature Number') % ylabel('Out-of-Bag Feature Importance') % % idxvar = find(b.OOBPermutedPredictorDeltaError>0.5) % idxCategorical = find(isCategorical(idxvar)==1); % % finbag = zeros(1,b.NTrees); % for t=1:b.NTrees % finbag(t) = sum(all(~b.OOBIndices(:,1:t),2)); % end % finbag = finbag / size(X,1); % figure % plot(finbag) % xlabel('Number of Grown Trees') % ylabel('Fraction of In-Bag Observations') % % b5v = TreeBagger(150,X(:,idxvar),Y,'Method','R',... % 'OOBPredictorImportance','On','CategoricalPredictors',idxCategorical,... % 'MinLeafSize',5); % figure % plot(oobError(b5v)) % xlabel('Number of Grown Trees') % ylabel('Out-of-Bag Mean Squared Error')