摘要
在治疗乳腺癌的药物研发过程中,通常采用建立化合物活性预测模型的方法来筛选潜在活性化合物。针对传统的回归预测模型对化合物活性预测的效果较差的问题,提出了一种基于粒子群算法优化的化合物活性预测模型,分别采用粒子群算法对SVR模型、Random Forest模型、XGBoost模型和LightGBM模型进行优化,对比分析优化前后的均方误差、平均绝对误差、拟合度等评价指标。结果表明,粒子群算法优化能够带来模型预测性能的提升,优化后的LightGBM模型预测效果更好,可为其他回归预测模型的优化提供方法参考。
In the process of drug development for the treatment of breast cancer,the method of building compound activity prediction models is usually used to screen potentially active compounds.To address the problem of poor prediction effect of traditional regression prediction models,a compound activity prediction model optimized based on particle swarm algorithm is proposed,and the particle swarm algorithm is used to optimize the SVR model,Random Forest model,XGBoost model and LightGBM model,respectively,to compare and analyze the evaluation of mean square error,mean absolute error and goodnessof-fit before and after optimization.The results show that the optimization of the particle swarm algorithm can improve the prediction performance of the models,and the optimized LightGBM model has better performance,which can provide a reference method for the optimization of other regression prediction models.
作者
王江翔
肖清泉
WANG Jiangxiang;XIAO Qingquan(College of Big Data and Information Engineering,Guizhou University,Guiyang 550025,China)
出处
《智能计算机与应用》
2023年第7期45-52,共8页
Intelligent Computer and Applications
基金
贵州省留学回国人员科技活动择优资助项目([2018]09)
贵州省高层次创新型人才培养项目([2015]4015)
贵州省研究生科研基金([2020]035)
贵州大学智能制造产教融合创新平台及研究生联合培养基地建设项目(2020520000-83-01-324061)。
关键词
粒子群算法
回归预测模型
化合物活性
particle swarm algorithm
regression prediction model
compound activity
