D.P is data points. T and P were temperature and pressure, respectively.

This model was compared with some other corresponding methods from literatures and the results were listed in Table 2. Based on GCM, Barati-Harooni et al.⁵¹ constructed a model to predict the density of ILs at variable temperature with AARD of 0.98 %. Lazzús⁵² and Chen et al.¹⁶ have made a good prediction for the density of ILs, in which the AARD value is about 0.73 % and 0.49 %. For the QSPR model of Lazzús⁵³, a moderate prediction result is obtained (AARD =2 %). Obviously, the statistical results of these models are good. Although the AARD (0.48 %) andR ² (0.9966) of this work are slightly worse than those of prior studies,⁵⁴ the amount of ILs datasets in this paper is the larger. At the same time, a large number of ILs datasets with different structures are used for modeling in our work, and the satisfactory prediction results further confirm the robustness and reliability of the ρ (T, P ) model.