Norm Indices-Driven Robust QSPR Model for Mining Temperature-Dependent Properties of Ionic Liquids

As a green solvent, temperature-dependent properties of ionic liquids (ILs) are important for their application. The large number of ILs makes experimental measurements of temperature-dependent properties nearly impossible to complete, which drives the development of models for filling in the data gaps. In this work, f(T,I)-QSPR models for refractive index (nD), heat capacity (Cp), surface tension (γ), and thermal conductivity (λ) are developed based on norm indices (Is) and prescreened extensive data sets. Importantly, leave-one-ion-out cross-validation (LOIO-CV) is employed to assess the stability of the QSPR models, ensuring the robust prediction of ILs with new cations and anions. The superiority of the four models is also confirmed by various evaluation means, including external validation, Y-random analysis, and the comparison with results from the literature. Given the satisfactory accuracy, these models will be useful for the rapid screening of IL properties data for the design of environmentally friendly ILs.