微型电脑应用

为了促进应用型科技成果的转化、提升科技资金投资的精度,对科技成果的社会效益预测方法进行了研究,构建了成果预测模型。该模型基于自组织神经网络在训练过程中引入神经元间的"竞争—合作"机制,解决了传统的神经网络对于高特征维度训练任务适应性差的难题;此外,该网络引入了邻域函数保存神经元间的拓扑关系,从而保证网络在训练过程中误差的稳定性;在模型特征向量的选取上,综合考虑应用型成果转移过程中科技成果项目本身的创新性、承担科技成果项目企业的能力、科技项目的经济可行性和成果项目管理团队管理水平等多个因素,构建了指标筛选体系。搜集了2013—2019年间724个科技项目的数据进行算法的仿真。仿真结果表明,与模糊评价算法相比,所提出的模型预测误差可以降低6.41%。

In order to promote the transformation of applied scientific and technological achievements and improve the accuracy of scientific and technological capital investment, the paper studies the social benefit forecasting methods of scientific and technological achievements, and constructs an achievement forecasting model. This model is based on a self-organizing neural network that introduces a competition-cooperation mechanism between neurons in the training process, to solve the problem of poor adaptability of traditional neural networks to training tasks with high feature dimensions. What is more, the network introduces a neighborhood function, saves the topological relationship between neurons, to ensure the stability of the error of the network during the training process. In the selection of the model feature vector, the innovation of the scientific and technological achievement project itself in the process of application-oriented achievement transfer is comprehensively considered. In addition, the scientific and technological achievement project is undertaken. As a result, an index screening system is established, it contains many factors such as the ability of the company, the economic feasibility of scientific and technological projects, and the management level of the project management team. Finally, the paper collected data from 724 science and technology projects from 2013 to 2019 and performed algorithm simulation. The simulation results show that the prediction error of the proposed model can be reduced by 6.41% compared with the fuzzy evaluation algorithm.