Scholarlink

<p>Using fluorescent organic dyes in various applications, such as OLEDs, optoelectronics, sensing, light harvesting, dye-sensitized solar cells, biomedical imaging, and pharmaceuticals, highlights the significance of accurately and quickly predicting their Stokes shift. For this, we proposed 15 regression models to estimate the Stokes shift of dyes efficiently and cost-effectively. The dataset of 3066 fluorescent organic materials used for training the models provides a robust foundation for their evaluation. The performance metrics, including <i>R</i><sup>2</sup>, MAE, and RMSE, were appropriately utilized to assess the models' accuracy. Based on the test scores, the Quadratic Support Vector Machine (QSVM) emerged as the top-performing model, demonstrating excellent prediction capabilities with an <i>R</i><sup>2</sup> of 80.40%, RMSE of 24.83&#xa0;nm, and MAE of 17.28&#xa0;nm. This model's accuracy was further validated using experimental data on three Coumarin dyes, where the Medium Gaussian Support Vector Machine (MGSVM) model showcased its superiority in predicting the Stokes shift. The comparison of relative errors between the MGSVM model and synthesized materials emphasizes the reliability of the proposed approach. The relative errors, ranging from 6.04 to 13.53&#xa0;nm, demonstrate the potential of these regression models to outperform current methods. In summary, this investigation addresses the challenges of directly measuring the Stokes shift of fluorescent materials utilizing solute and solvent structures. The proposed models provide quick and accurate predictions and offer a cost-effective solution for screening a wide range of dyes. This research paves the way for researchers to gain valuable insights into materials beforehand and expedite the discovery of new compounds.</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

Leveraging optimized machine learning regression models for measuring the Stokes shift of fluorescent organic dyes

  • 期刊论文

  • 发表日期：

    2025年8月13日

查看全文
  • Kapil Dev Mahato,
  • Uday Kumar

摘要

Using fluorescent organic dyes in various applications, such as OLEDs, optoelectronics, sensing, light harvesting, dye-sensitized solar cells, biomedical imaging, and pharmaceuticals, highlights the significance of accurately and quickly predicting their Stokes shift. For this, we proposed 15 regression models to estimate the Stokes shift of dyes efficiently and cost-effectively. The dataset of 3066 fluorescent organic materials used for training the models provides a robust foundation for their evaluation. The performance metrics, including R2, MAE, and RMSE, were appropriately utilized to assess the models' accuracy. Based on the test scores, the Quadratic Support Vector Machine (QSVM) emerged as the top-performing model, demonstrating excellent prediction capabilities with an R2 of 80.40%, RMSE of 24.83 nm, and MAE of 17.28 nm. This model's accuracy was further validated using experimental data on three Coumarin dyes, where the Medium Gaussian Support Vector Machine (MGSVM) model showcased its superiority in predicting the Stokes shift. The comparison of relative errors between the MGSVM model and synthesized materials emphasizes the reliability of the proposed approach. The relative errors, ranging from 6.04 to 13.53 nm, demonstrate the potential of these regression models to outperform current methods. In summary, this investigation addresses the challenges of directly measuring the Stokes shift of fluorescent materials utilizing solute and solvent structures. The proposed models provide quick and accurate predictions and offer a cost-effective solution for screening a wide range of dyes. This research paves the way for researchers to gain valuable insights into materials beforehand and expedite the discovery of new compounds.