基于BP神经网络的高校3种专利授权量预测研究  

The Forecast Research of Three Invention Patents Authorization of Colleges Based on BP Network

作　　者：臧西杰[1] 禹建丽[1] 周瑞芳[1] 

机构地区：[1]中原工学院,郑州450007

出　　处：《中原工学院学报》2009年第6期35-37,44,共4页Journal of Zhongyuan University of Technology

基　　金：河南省科技攻关项目(072102260020)

摘　　要：以1985—2007年授权量的数据为基础,运用BP神经网络建立预测模型,并结合Matlab工具箱对算法进行了实现.通过设置Matlab神经网络训练函数的最大训练次数300、收敛精度1e—6及最小梯度1e—6,合理确定了最优专利授权量的BP神经网络模型.仿真验证表明:该模型预测1990—2007年专利授权量时,平均绝对百分误差MAPE值为6.87%,逼近效果非常好.最后,利用该模型对未来5年的高校专利授权量进行了预测.The BP network models were build for time series forecasting since the complexity and nonlinear of data. The optimum network was build by trainning times setting as 300, precision and the minimum gradient setting as le-6. The simulation result shows that the forecast has high accuracy, and the mean absolute perecent error was only 6.87 %. The five-year-forecast return was made by using the optimum network.

关 键 词：高校 专利授权量 时间序列分析 BP 

分 类 号：O236[理学—运筹学与控制论]