Generalized PLS regression forecast modeling of warship equipment maintenance cost

Aiming at the small sample, more latent variables and the multicollinearity among them in the forecast modeling of warship equipment maintenance cost, a method of improving the generalization ability of PLS was presented, which was on the base of the partial least squares(PLS) regression with shrink-magnifying, and extended the shrinking factor more by shrinking or magnifying the inputs of different sample to different extent, in which the cross training between training set and testing set was implemented. Further more, the foregoing modeling process was applied to the forecast modeling of warship equipment maintenance cost, in which the genetic algorithm was used to seek the best shrinking factor vector. Finally, by comparing with the PLS which distills one, two and three principal components and PLS with shrink-magnifying approach, the method presented in this paper demonstrates the best.