Retrieve variables used for prediction in data analysis solutions/models

[bburlacu]

Some optimization methods have the ability to perform automatic feature selection, producing models that use a subset of the allowed input variables. It would be useful in this case to be able to interogate each model about the variables used for prediction. The collection of allowed input variables would be returned by default, while symbolic data analysis models for example would return the collection of variables present in the tree.

[bburlacu]

r13826: Decided to implement this functionality at the level of the DataAnalysisSolution (and not the model), since this way we have access to the problem data and can provide a default implementation of the new interface method so that this change will have no negative repercussions.

[bburlacu]

r13921: Revert changes to DataAnalysisSolution and IDataAnalysisSolution, and implement the desired properties in model classes that implement IDataAnalysisModel, IRegressionModel and IClassificationModel.

[bburlacu]

r13922: Forgot to add VariablesUsedForPrediction get method for GaussianProcessModel.

[mkommend]

r13941:

• Base classes for data analysis, classification, and regression models
• Added target variable to classification and regression models
• Switched parameter order in data analysis solutions (model, problemdata)

[gkronber]

r13992: set default target variable to string.Empty in AfterDeserialization hook for regression and classification models and changed a few ctor calls to correctly pass the target variable

[gkronber]

Some models are used both for classification and regression problems. They are now all derived from ClassificationModel base class (e.g. NeuralNetworkModel)

[gkronber]

r13993: minor changes in time series models and fix build fail

[gkronber]

I reviewed r13826, r13921:13922 and r13941. Changes seem to be ok. (I think elastic-net branch needs to be updated)

Could you please review r13992:13993. Please release if these changes are ok.

[mkommend]

Reviewed r13992:r13993.

[mkommend]

r14000: Removed default ctor arguments for the target variable in regression and classification models.

[mkommend]

Could you please review r14000. Please release if these changes are ok.

[gkronber]

Reviewed r14000.

[mkommend]

r14027: Merged r13826,r13921, r13922, r13941, r13992, r13993, r14000 intos table.