Borrar filtros
Borrar filtros

Error (misclassification probability or MSE)

5 visualizaciones (últimos 30 días)
Stephen Gray
Stephen Gray el 4 de Jun. de 2024
Comentada: Stephen Gray el 10 de Jun. de 2024
Hi all. I am trying to use the Error (misclassification probability or MSE) function. In the help documentation it shows an example of useage as err = error(B,TBLnew,Ynew). I tried this with err=error(Mdl,Inps,Outs) with Mdl being the LSBOOST ensemble, Inps being the matrix of vars and Outs being the target. If I try this I get 'Error using error Too many output arguments.' . I presume I'm missing something here but can't seem to find a an example.

Respuesta aceptada

Simar
Simar el 10 de Jun. de 2024
Editada: Simar el 10 de Jun. de 2024
Hi Stephen,
As per my understanding you are encountering an issue with error function when trying to evaluate the performance of an LSBoost ensemble model in MATLAB. The error message "Too many output arguments" indicates a mismatch between function's expected output and how it is being called in the code.
For ensemble models like LSBoost, MATLAB uses loss function to compute the prediction error, rather than error. Evaluating an ensemble model, one can measure its performance by calculating the loss, which can represent misclassification probability for classification problems or mean squared error (MSE) for regression problems. Please refer to “Loss” documentation link- https://www.mathworks.com/help/stats/regressionneuralnetwork.loss.html?searchHighlight=loss&s_tid=srchtitle_support_results_2_loss
Here is how one can use loss function with an LSBoost ensemble model:
1. For Classification:
If target variable Outs is categorical (for a classification problem), calculate the classification error (misclassification probability) as follows:
>> L = loss(Mdl, Inps, Outs, 'LossFun', 'classiferror');
In this context, Mdl is the trained LSBoost model, Inps are the input variables, and Outs are the actual class labels.
2. For Regression:
If dealing with regression problem (where Outs is a continuous variable), calculate the mean squared error (MSE) as follows:
>> L = loss(Mdl, Inps, Outs);
Here, default loss function for regression problems is mean squared error, so do not need to specify 'LossFun' unless want to use a different metric.
3. General Usage
>> L = loss(Mdl, Inps, Outs, 'LossFun', lossFunction);
  • Mdl is the LSBoost model.
  • Inps is the matrix of input variables.
  • Outs is the vector of actual outputs (targets).
  • lossFunction specifies the loss function to use, such as 'classiferror' for classification or 'mse' for regression.
Ensure that Inps and Outs are correctly formatted for model's expectations.
Hope it helps!
Best Regards,
Simar

Más respuestas (0)

Productos


Versión

R2024a

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by