基于权重初始化-多层卷积神经网络滑动窗口融合的高等教育办学规模预测算法

高等教育办学规模预测主要有自回归法、支持向量机法等;这些具体方法在探索高等教育办学规模方面都取得了一定成果。但都不能较好地揭示高等教育办学规模的演化规律。鉴于此,本文引入深度学习模型对高等教育办学规模进行建模并预测。本文首先从网络优化角度出发,通过优化方法提高网络的非线性建模能力,提出了一种新的网络权重初始化方法,增加了神经网络处理不同视觉任务的潜力。同时,本文提出了适合高等教育办学规模数据演化规律的异构多列卷积神经网络。最后,利用本文提出的自适应的滑动窗口融合机制完成高等教育办学规模预测。利用本文方法对某区域高等教育办学规模进行实例分析,结果表明,本文所提方法能够较好地给出相应演化规律。

The scale of running higher education is mainly based onautoregressive method,support vector machine method decrease method.However,they cannot better reveal the evolution of the scale of higher education.So,this paper introduces a new deep learning model to deeply model and predict the scale of higher education.Firstly,from the perspective of network optimization,this paper proposes a new network weight initialization method,which alleviates the problem that the initialization theory of existing deep learning is limited by the nonlinear unit type,which increases the potential of neural network to deal with different visual tasks.Moreover,this paper finds a convolutional neural network suitable for the evolution of the scale data of higher education.Finally,using the method proposed in this paper to complete the forecast of the scale of higher education.The method of this paper is used to analyze the scale of higher education in a certain region.The results show that the proposed method can give a corresponding evolution law.