Optimization of Virtual Samples Number Based on Information Entropy

Vessel noise sources identification system can be modeled by identification system modeling methods such as Artificial Neural Network. The difficulty of experiment and measurement make us unable to get enough real samples to guarantee the quality of identification system modeling. This paper proposes a general method to regenerate virtual samples for identification system modeling based on bootstrap and an optimization rule of virtual samples number as "Same Weightiness, Same Information" based on information entropy in order to model identification system more efficiently. The results of three experiments validate optimization rule of virtual samples number in regeneration virtual samples and identification system modeling.