Quantify extrapolation of new samples given a training set
R
The applicable package helps determine when model predictions may be unreliable by measuring how much new data points differ from the training set. It implements “applicability domain” methods to detect potential extrapolation.
This package is valuable for identifying when predictions should be treated with skepticism because the input data falls outside the model’s training domain. It provides multiple methods for both binary and continuous data to quantify the degree of extrapolation. This is particularly useful in fields like chemistry and other scientific domains where prediction reliability is critical.
