基于改进PSO-BP神经网络算法的半导体材料带隙宽度预测  

Prediction of Band Gap Width of Semiconductor Materials Based on Improved PSO-BP Neural Network Algorithm

作　　者：肖斌[1] 胡国梁 XIAO Bin;HU Guoliang(School of Computer Science,Southwest Petroleum University,Chengdu Sichuan 610500,China)

机构地区：[1]西南石油大学计算机科学学院,四川成都610500

出　　处：《电子器件》2022年第2期282-286,共5页Chinese Journal of Electron Devices

摘　　要：半导体材料的带隙宽度对其性能有重要影响,准确预测带隙宽度对半导体材料的研究具有重要意义。通过密度泛函理论计算半导体材料带隙宽度通常需要耗费大量的时间且预测精度较低,因此建立了一种基于统计学方法和改进PSO-BP神经网络算法的半导体材料带隙宽度预测模型,用于提高带隙值的预测精度。该模型先通过统计学的方法对半导体材料带隙宽度数据集的特征属性进行分析和选择,而后利用改进的PSO-BP神经网络算法挖掘特征属性与带隙值之间隐含的数学关系。实验结果表明,该预测模型相比于对照模型的均方误差降低了25%以上,可靠性达到了75.15%,可广泛应用于需要大量预测半导体材料带隙宽度的场合。Band gap width of semiconductor materials has an important influence on its performance.Accurate prediction of band gap width is of great significance to the research.Density functional theory usually takes much time to calculate band gap width,and the prediction accuracy is low.Therefore,a prediction model of band gap width based on statistical methods and an improved PSO-BP neural network algorithm was established to improve the prediction accuracy.The characteristic attributes of the semiconductor material data set was first analyzed and selected through statistical methods,and then the improved PSO-BP neural network algorithm was used to mine the implicit mathematical relationship between characteristic attributes and band gap.Experimental results show that,compared with the control model,the error of prediction was reduced by more than 25%,and the reliability reached 75.15%.This model can be widely used when predicting band gap width of many semiconductor materials.

关 键 词：材料带隙宽度 材料理论计算 半导体材料 粒子群优化算法 BP神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TN304[自动化与计算机技术—控制科学与工程]