基于PCA-BP模型的高校教师职称评审预测

Professional Titles Evaluation Forecasting Based on PCA-BP

针对高校教师职称评审问题,提出基于PCA-BP的评审预测模型。采用主成分分析法对评审指标数据进行降维处理,选取保留原始指标信息的89.01%的四个主成分作为BP网络的输入,这样不仅减少了网络的输入维数,减小了网络训练规模,而且消除了各指标间的相关性,改善了网络的训练效率,提高了预测精度。利用Matlab软件对某高校2012年副教授评审实际数据的进行实例分析和仿真,并用该组数据比较该方法与典型BP网络的预测效果,结果表明该方法明显优于BP网络,完全能够满足职称评审预测的要求。

This paper presented a professional title forecast model based on PCA and BP neural network.The principal component analysis is used to reduce the dimension of assessment indicators data,and four main components retaining 89.01% of the original information as the BP network input are selected.It not only reduces the number of input variables and the size of NN,but also irrelevance among variables and improve the efficiency of the network＇s training and the prediction accuracy.An example for a professional title is given.Simulation results show that the PCA –BP neural network model is superior to traditional BP neural network,and meet the requirement for professional title prediction.