3.2.2 Internal-validation and External-validation
The Cross-validation calculated values versus experimental values of log η data were showed in Figure 11 (a). The statistical parameters for validation are listed in Table 3. The Figure 11 (a) shows that the experimental value is close to the log η value calculated by Cross-validation. Table 3 shows that the Q2 (0.9668), AARD (3.95%) andRMSE (0.1276) are acceptable, which suggests the good stability and robustness of this η model.
7432 data points were randomly divided into the training set (5874 data points) and the testing set (1468 data points). The calculated values for the training set and the testing set with the experimental values were showed in Figure 11 (b). The performance parameters for external validation results are listed in Table 3. External validation results showed that the R 2 of training set and testing set were 0.9535 and 0.9607, and the AARD were 5.051 % and 4.743 %, respectively. These result showed that the QSPR model based on norm descriptor was stable and has good predictive ability for ηprediction of ILs with diverse structures.