Scholarlink

The study utilized the PM6 method from the Ampac10 software package in conjunction with Codessa 3.3 to calculate 958 descriptors for 275 compounds. Through preprocessing and feature variable screening based on genetic algorithm, 48 significant feature descriptors influencing Global Warming Potential (GWP) were identified. An 85%–15% split was randomly selected for the training and prediction sets, respectively, to establish a QSPR model for GWP using neural network training. The model’s R2 was 0.9142, indicating a correlation coefficient R of 0.9561, which suggests good predictive ability. Additionally, the residual diagram indicates no systematic error in the model establishment process. In conclusion, the combination of the PM6 method in Ampac10, Codessa 3.3 software, genetic algorithm, and neural network training can become a new, simple, and accurate method for predicting GWP.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Research on GWP Prediction Using Genetic Algorithm and Neural Network

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Dongwei Sun,
  • Weipeng Lai,
  • Yiding Ma,
  • Tao Yu,
  • Yingzhe Liu,
  • Nian Tang,
  • Li Li

摘要

The study utilized the PM6 method from the Ampac10 software package in conjunction with Codessa 3.3 to calculate 958 descriptors for 275 compounds. Through preprocessing and feature variable screening based on genetic algorithm, 48 significant feature descriptors influencing Global Warming Potential (GWP) were identified. An 85%–15% split was randomly selected for the training and prediction sets, respectively, to establish a QSPR model for GWP using neural network training. The model’s R2 was 0.9142, indicating a correlation coefficient R of 0.9561, which suggests good predictive ability. Additionally, the residual diagram indicates no systematic error in the model establishment process. In conclusion, the combination of the PM6 method in Ampac10, Codessa 3.3 software, genetic algorithm, and neural network training can become a new, simple, and accurate method for predicting GWP.