ChemFluor

We establish a machine learning-based method to predict emission/absorption wavelength and PLQY of organic fluorescent materials.A platform has been establised for experimenters to use, as well as used for potential high-throughput screening.<br><b>[1]</b><i>The ChemFluor_v0.1.zip</i> is the platform based on python, which contain trained models, can be used for the prediction directly.<b>[2]</b><i>Fingerprints_for_prediction.zip</i> is the fingerprints used in our work.<b>[3]</b><i>Materials_Real-World_Problem.zip</i> is the molecules collected from recent published work and TD-DFT benchmark studies, which can be seen as real world problem. The molecules are stored in the form of SMILES. <b>[4]</b><i>Alldata_SMILES_v0.1.xlsx</i> contains all the molecules in our dataset as well as the references.<br><b>[5]</b>ML-models we used in our paper for real world problems have been saved and uploaded, as<i> model_in_paper.zip.</i><b>[6]</b><i>Molecule-based_partition_withFP.zip </i>contain the training set and test set mentioned in our updated manuscript. We split our dataset based on molecules but not data-points in this file. <i><br></i><i>-----</i><i>update in 2020.07.21</i>[1] QY regressor model have been supported.我们构建了一种基于机器学习的方法，用于预测有机荧光材料的发射/吸收波长与光致发光量子产率（Photoluminescence Quantum Yield, PLQY）。本研究搭建了可供实验人员使用的平台，同时可用于潜在的高通量筛选。<br><b>[1]</b><i>ChemFluor_v0.1.zip</i> 为基于Python开发的平台，内置训练完成的模型，可直接用于预测任务。<b>[2]</b><i>Fingerprints_for_prediction.zip</i> 包含本研究中使用的分子指纹。<b>[3]</b><i>Materials_Real-World_Problem.zip</i> 收录了从近期发表文献及含时密度泛函理论（Time-Dependent Density Functional Theory, TD-DFT）基准测试研究中获取的分子，可作为真实世界测试任务使用，所有分子以SMILES（Simplified Molecular Input Line Entry System, SMILES）格式存储。<b>[4]</b><i>Alldata_SMILES_v0.1.xlsx</i> 收录了本数据集的全部分子及相关参考文献。<b>[5]</b> 本研究论文中用于真实世界任务的机器学习（Machine Learning, ML）模型已保存并上传，文件名为 <i>model_in_paper.zip.</i><b>[6]</b><i>Molecule-based_partition_withFP.zip</i> 包含本研究修订版手稿中提及的训练集与测试集，该文件按分子而非数据点对数据集进行划分。<i><br></i><i>-----</i><i>2020年7月21日更新</i> [1] 新增量子产率（Quantum Yield, QY）回归模型支持。