Abstract: Aiming at the high training complexity of neural network situation prediction model, a situation prediction method based on improved convolution neural network is proposed. Combined with the advantages of depth-wise separable convolution and factorization into smaller convolution, a new model of improved convolution neural network security situation based on composite convolution structure is proposed, and the mapping of situation elements and situation values are realized. The experimental simulation results show that the method obviously reduces the time complexity and the prediction time, improves the prediction accuracy compared with the existing typical prediction methods.

Key words: situation prediction, neural network, convolution neural network, compound convolution structure

摘要： 针对神经网络态势预测模型训练复杂度高的问题，提出了一种基于改进卷积神经网络的态势预测方法。结合深度可分离卷积与分解卷积技术的优点，提出了一种基于复合卷积结构的改进型卷积神经网络安全态势预测模型，实现了态势要素和态势值的映射。实验仿真结果证明，相比于已有的典型预测方法，该方法明显降低了复杂度，减少了预测时间，并提升了预测准确率。

关键词: 态势预测, 神经网络, 卷积神经网络, 复合卷积结构