LocalModel

Declaration:

TMemBasedLearner.LocalModel: TMBLLocalModel;

The property LocalModel controls how the target values are estimated from the k nearest neighbors of the model data. Currently the following four modes of estimation are available: mblLMAverage The result is the mean of the target variables of the k nearest neighbors. mblLMWgtGauss The result is the weighted mean of the target variables of the k nearest neighbors. The weights depend on the distances of the respective neighbors (bell shaped decay whose width is controlled by the standard deviation of the k distances). mblLMMajority The result is determined by the majority of the target variables of the k nearest neighbors. Please note that in the case of non-categorical target variables the majority is determined by evaluating the density of the target values. mblLMRegress The result is estimated by a linear regression model established by the k nearest neighbors. Please note that linear regression model need at least one data point more than variables, which means the NumNearestNb must be greater than the number of independent variables. For stable linear models the number of neighbors should be at least twice the number of independent variables.