机器学习预测无机金属卤化钙钛矿带隙
首发时间:2021-02-24
摘要:本文从开源材料数据库Material Project中获取无机金属卤化钙钛矿数据115条,并利用随机森林(RF),以及梯度增强回归树(GBRT)算法对钙钛矿材料的带隙进行预测。两种模型中,梯度增强回归树模型的预测精度更高最佳性能可达到平均预测误差为0.25eV。对GBRT模型的输入特征进行分析后,发现B位离子电负性,最高占用轨道能级和晶格常数a三个特征对与带隙预测最为重要,并且与带隙预测值呈负相关关系。
关键词: 无机金属卤化钙钛矿 机器学习 梯度增强回归树 极端回归树
For information in English, please click here
Using machine learning to predict bandgap of inorganic metal halide perovskite
Abstract:In this paper,we obtain a dataset consisted of 115 inorganic metal halide perovskites from Material Project, and train two machine learning models(ET and GBRT) on it to predict the bandgaps of inorganic metal halide perovskites. Among them, GBRT has a better performance and its average prediction error is 0.25eV. After analyzing the input features of GBRT, it is shown that B_Electronegativity, HOMO and lattice constant a are the most three important features for bandgap prediction, and they all have relatively negative correlations with the value of bandgap prediction.
Keywords: Inorganic metal halide perovskite, Machine learning, GBRT, ET?????
基金:
引用
No.****
动态公开评议
共计0人参与
勘误表
机器学习预测无机金属卤化钙钛矿带隙
评论
全部评论0/1000