Research on The Design of BP Neural Network for Calculating Combat Effectiveness Index
The quantitative study of combat effectiveness index is crucial for the informationization construction of the military.In response to the shortcomings of the current research method for combat effectiveness index,such as being greatly influenced by subjective factors,more qualitative analysis,less quantitative research,insufficient method generalization,etc.the importance and feasi-bility of using BP neural network as the"black-box model"to calculate the combat effectiveness index are reiterated.This method takes the indicators that affect the combat effectiveness index as input and the values of the combat effectiveness index as output,through the training method of er-ror back-propagation and gradient descent,automatically learns the complex function relationship between output and input.In addition,a set of effective optimization methods for BP neural net-works is proposed to address the current issues such as limited research on BP neural network structure optimization,simple structure of BP neural network,and inability to guarantee the per-formance and stability.This method optimizes the BP neural network through three approaches:optimizing the number of neurons,hiding layers and training methods.Simulation experiments show that all three optimization methods can improve the performance of BP neural network.The average prediction error of the best model with the same optimization method is about 4％,7％and 6％lower than that of other models respectively.
NETL
NSTL
万方数据
