MARSplines
MARSplines (Multivariate Adaptive Regression Splines) - Nonparametric statistical regression. The method does not assume a particular form of dependency between predictors and explanatory variables, but instead selects specific data.
MARSplines General Model: Y = β 0 + ∑ m = 1 M β m h m ( X ) {\displaystyle y=\beta _{0}+\sum \limits _{m=1}^{M}\beta _{m}h_{m}(X)}
where:
The method requires prior design of an appropriate set of base functions, which may in particular include the values of individual predicates, their polynomials, or modeling interactions between several predicates.
The algorithm searches the space of the possible weighted average of the values of the base functions and iteratively selects an increasingly large number of them using the criterion of minimizing the squares of errors (least squares method). Bibliography
wiki
Comments
Post a Comment