基于神经网络的高校科研项目前景预测系统研究与设计
2021-03-24王汉元姜英
王汉元 姜英



摘 要:为了对高校科研项目前景进行有效预测,进一步提升科研成果转化率,本文提出了基于神经网络结合科研项目共享的网络平台,构建可靠的科研项目前景预测系统。通过将各类大量科研项目的特征参数化,利用BP(反向传播)神经网络的非线性映射能力、自适应能力和对离散数据的泛化能力生成模型后,用于从多维度对一个新项目生成可靠的前景预测。实验表明,BP神经网络算法在大量学习经过预处理的样本后所产生的预测结果是具有较高准确性的,且能够随着新的样本输入不断更新和适应,因此该方法具有较强的可行性。
关键词:高校科研项目;前景预测;神经网络
中图分类号:TP183 文献标识码:A
Abstract: This paper proposes to build a reliable prospect forecast system for scientific research projects in order to effectively predict prospects of scientific research projects in universities and further improve conversion rate of scientific research results. The system is based on neural networks and a network platform shared by scientific research projects. A model is generated by parameterizing characteristics of a large number of various scientific research projects, using the non-linear mapping ability, adaptive ability and generalization ability of BP (Back Propagation) neural network. Then, the model is used for reliable forecast of a new project from multiple dimensions. Experiments show that the prediction results, produced by the BP neural network algorithm after learning a large number of pre-processed samples, have high accuracy, and can be updated and adapted with new sample input. Therefore, this method is highly feasible.
Keywords: university scientific research projects; prospect prediction; neural network
1 引言(Introduction)
在我国,科研成果转化率低已是老生常谈的话题,2018年4月18日国务院常务会议确定了“加大对科技成果转化的政策激励”的政策方针并提出了若干办法。但由于科研评价机制不合理、选题立项重复度高、科研资金管理存在漏洞[1]等问题,导致科研成果转化率低的现象至今依然存在。提高科研成果转化率,企业的参与是至关重要的,然而企业为了规避风险很注重项目的成熟度,偏向于投资前景已清晰的项目。这样便导致很多优秀的项目由于找不到投资者止步在“原始创新成果”阶段,失去了产业化、商品化的机会。而各个高校的科研项目展示页面更注重展示本校科研项目的创新性和技术高度,而忽略了展示方案的实现成本及产业化的难度[2],这让企业难以理解其价值。……
