As memory based learners (also known as kNN) heavily rely on data (the data is the model) the central hub of TMemBasedLearner is the property ModelData which contains all the training data. The data is organised as TDataTable, with the rows being the objects and the columns being the variables.
The type of the variables is encoded as column attribute (do not mix up the variable type with the measurement scale) - which means that the column attributes are not available for free usage. TMemBasedLearner uses them internally and make them accessible by the property VarType. All other properties and methods of the ModelData variable can be used by the user.
|