LibSVM训练错误 [英] LibSVM training error

问题描述

出于预测目的，我正在使用Java版本的libsvm(回归). 在训练完我的数据集之后，生成的模型显示了支持向量，但是没有显示训练错误率. 我想知道是否可以找到我的训练集的训练错误?有什么我可以调用的函数，或者可以用来查找它的类属性吗?

I’m using java version of libsvm (regression) for prediction purposes. After training my data set, the generated model shows the Support Vectors, but no indication on the training error rate. I would like to know if it's possible to find the training error of my training set? Is there any function I can call, or a class attribute I can use to find it?

谢谢

推荐答案

您将不得不从训练集中创建一个svm_problem，并在该集合上调用svm_predict(..)-然后计算MSE(均方误差)-但是请注意，与神经网络相比， LibSVM在回归数据集上的表现相当差..

You will have to create a svm_problem from your training-set and call svm_predict(..) on that - then compute the MSE (mean squared error) - but note that LibSVM performs quite poor on regression datasets compared to neural networks.

这篇关于LibSVM训练错误的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！