|
|
|
|
|
|
 |
 |
|
Support Vector Regression and Other Prediction Methods: A Competition with Mathematica
|
|
|
|
|
|
|
|
|
|
|
|
2017-01-09
|
|
|
|
|
|
We apply eight prediction methods to eleven data sets and compare the prediction capabilities of the various methods. The methods are polynomial regression, support vector regression, local regression, and the five methods provided by Predict: linear regression, neural network, Gaussian process, nearest neighbors, and random forest. For support vector regression and local regression, we have written our custom programs. The method of support vector regression is derived and explained at some length. The fits are validated by custom cross-validation. The winner of our competition is support vector regression; then comes local regression. For support vector regression, we consider separately Gaussian and polynomial kernels; polynomial kernels were better in our study.
|
|
|
|
|
|
|
|
|
|
|
|
support vector regression, prediction, competition, polynomial regression, local regression, Predict, linear regression, neural network, Gaussian process, nearest neighbors, random forest, cross-validation, Gaussian kernel, polynomial kernel, machine learning
|
|
|
|
|
|
| SupportVectorRegression1.zip (4 MB) - ZIP archive | | SupportVectorRegression2.zip (6.7 MB) - ZIP archive | | SupportVectorRegression3.zip (10.1 MB) - ZIP archive | | SupportVectorRegression4.zip (2.3 MB) - ZIP archive |
|
|
|
|
|
|
|
| | | | | |
|
For the newest resources, visit Wolfram Repositories and Archives »
