Where can I find information for variable descriptions I have been working in R the last few weeks and have been tinkering with forecasting/predicting values for the financial data.
Is there a good place to find out what the different variables represent? Such as m in caret package, size. C, degree, sigma, and scale. 
For example I am trying to use the Support Vector Machines models in caret but they require some input parameters for the different types.
Support Vector Machines
type, package, variables
svmLinear, kernlab, C
svmRadial, kernlab, C, sigma
svmRadialCost, kernlab, C
svmPoly, kernlab, degree, scale, C

I found the above in documentation regarding the SVM Models in the caret package. However, no explanation on what the variables mean....
Would there be a good place to figure out what these variables mean? Or are they already meant to be known by even novices? If so where can I find out about them?
I think it would be best if I knew what these were instead of just changing them randomly.
 A: The parameters you're describing are standard kernel parameters in SVM classification. I almost always forget what they mean too, but have found the LibSVM website to be a solid reference for this. From the site, linear is the cost parameter, which is appears in every kernel, and
$$\text{Linear: } u'*v$$
$$\text{Polynomial: } (\gamma*u'*v + coef_{0})^{degree}$$
$$\text{RBF: } exp^{(-\gamma*|u-v|^{2})}$$
$$\text{Sigmoid: } tanh(\gamma*u'*v + coef_{0})$$
Does that help?
A: All of the parameters required by caret models are "hyper parameters" that control the type of model fitted to the data.  It's a good idea to choose these parameters using cross-validation, which is what caret does.
In the case of the support vector machine, you don't really need to understand what these parameters represent, because you probably shouldn't choose them by hand.  Just try a reasonable range of values and select the best ones based on cross-validated error.  (hint: this is exactly what caret does by default!)
If you're dealing with time series data, note that caret isn't really setup to handle this situation by default, and you may need to construct your own mechanism for cross-validating time series data.
