Scholarlink

<p>Machine learning models can be applied for estimation of continuous manufacturing parameters in pharmaceutical processing of oral-solid formulations. Development of Quality by Design (QbD) has motivated the pharmaceutical sector to move towards continuous manufacturing by developing advanced computational models as well as analytical techniques. Despite the application of conventional methods, supercritical fluids (SCFs) have offered a new way for particle generation for nanonization and advanced manufacturing. In this work, the solubility of Letrozole was evaluated using temperature and pressure correlated to machine learning. KNN (K-Nearest Neighbors) and two boosted versions employing AdaBoost and bagging ensemble models are used. Golden eagle optimizer (GEOA) was applied as optimizer and tweaking hyper-parameters. Once the models were optimized, they were assessed using various performance metrics. The R-squared scores achieved were 0.9907 for KNN, 0.9945 for AdaBoost-KNN, and 0.9938 for Bagging-KNN. Among these, the AdaBoost-KNN model demonstrated the highest accuracy.</p>

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

Computational analysis on the influence of pressure and temperature on drug solubility in supercritical CO2 with machine learning and optimizer

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月17日

查看全文
  • Ahmad J. Obaidullah,
  • Wael A. Mahdi,
  • Adel Alhowyan

摘要

Machine learning models can be applied for estimation of continuous manufacturing parameters in pharmaceutical processing of oral-solid formulations. Development of Quality by Design (QbD) has motivated the pharmaceutical sector to move towards continuous manufacturing by developing advanced computational models as well as analytical techniques. Despite the application of conventional methods, supercritical fluids (SCFs) have offered a new way for particle generation for nanonization and advanced manufacturing. In this work, the solubility of Letrozole was evaluated using temperature and pressure correlated to machine learning. KNN (K-Nearest Neighbors) and two boosted versions employing AdaBoost and bagging ensemble models are used. Golden eagle optimizer (GEOA) was applied as optimizer and tweaking hyper-parameters. Once the models were optimized, they were assessed using various performance metrics. The R-squared scores achieved were 0.9907 for KNN, 0.9945 for AdaBoost-KNN, and 0.9938 for Bagging-KNN. Among these, the AdaBoost-KNN model demonstrated the highest accuracy.