Research on Threat Assessment evaluation model based on improved CNN algorithm

In view of the traditional Threat Assessment (TA) evaluation model can only consider a single threat target, and the accuracy of threat evaluation is poor, the application effect of improved CNN algorithm in Ta evaluation model is studied. This paper proposes a TA evaluation model based on the improved Convolutional Neural Networks (CNN) algorithm. The model uses the powerful feature extraction ability of convolutional neural network, adopts the concept of dual channel neuron, improves the structure of convolutional neural network, and reduces the number of network parameters and obtains the target classification features with multiple markers on the basis of retaining the full connection layer. On this basis, fuzzy mathematics is used to quantitatively describe the classification features of multi marker targets, to define the weight value of each feature of targets, and to evaluate the threat degree of multiple targets by Technique for Order Preference by Similarity to Ideal Solution (TOPSIS). The simulation results show that the model has fast convergence speed and accurate threat prediction ability, and can accurately obtain the threat ranking of multiple targets.

Automated summary

This paper investigates the application effect of the improved Convolutional Neural Networks (CNN) algorithm in the Threat Assessment (TA) evaluation model, which addresses the weaknesses of the traditional model in considering only a single threat target and having poor accuracy. The proposed TA evaluation model uses the powerful feature extraction ability of CNN, adopts the concept of dual channel neuron, improves the structure of CNN, reduces network parameters, and obtains target classification features with multiple markers while retaining the full connection layer. The model utilizes fuzzy mathematics and Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) to quantitatively describe the classification features of multi-marker targets, define feature weights, and evaluate the threat degree of multiple targets. Simulation results demonstrate that the model possesses fast convergence speed, accurate threat prediction ability, and the capability to accurately rank multiple targets by threat level.