How to predict output values of new data in stepwise regression?

Question

Ajay Goyal 2015-12-28

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/261785-how-to-predict-output-values-of-new-data-in-stepwise-regression

评论： Ajay Goyal 2015-12-29

Dear Friends, I have generated a stepwise quadratic model using stepwiselm. At start of running the code, I have divided my training and tested data. After training (generating the model) I want to calculated output for tested data. I am trying to use predict command but unable to execute it. Could you please help me (maybe with an example).

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Ajay Goyal 2015-12-29

Thank You Proff Brendan for your help. It worked. Yesterday I was doing a silly mistake. Thanks once again for enlighten me from your knowledge

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Brendan Hamm 2015-12-28

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/261785-how-to-predict-output-values-of-new-data-in-stepwise-regression#answer_204441

在 MATLAB Online 中打开

rng('default') % Reproducibility
n = 200; % 200 samples
% Generate some data which is quadratic
x = 4*(rand(n,1)-0.5); % Samples from U([-2 2])
y = 1.2 + 7.8*x -2.6*x.^2 + 5*randn(n,1);
pt = cvpartition(n,'Holdout',0.5); % Create partition
trainX = x(training(pt)); % Get training predictor data
trainY = y(training(pt)); % Get training response data
testX = x(test(pt)); % Get validation predictor data
testY = y(test(pt)); % Get validation response data
% Plot test data:
plot(trainX,trainY,'ko')
% Fit model and predict:
mdl = stepwiselm(trainX,trainY,'y ~ 1','Criterion','BIC','Upper','y~1+x1^5');
predY = predict(mdl,testX);
% Plot predictions against known response
plot(testX,predY,'r*')
hold on
plot(testX,testY,'bo')

2 个评论
显示无隐藏无

Brendan Hamm 2015-12-28

If this is not working you may have made your own function called predict which shadows the method. What is happening when predict is "not working"? Is it an error, and if so what does it say?

Are you perhaps looking at the documentation for a different predict (there are many for different classes).

Ajay Goyal 2015-12-29