2.3.1 SVM and MLR Models
QSAR models were constructed through Support vector machine (SVM)17 and multiple linear regression (MLR)16 machine learning algorithms which were established in libSVM software and Matlab software.
Through equation (2), we can acquire the SVM model’s square correlation coefficient (R2) which can represent SVM is suitable to predict compounds abilities.
\(\text{\ \ R}^{2}=\ \frac{\left(\overset{\overline{}}{l}\sum_{i=1}^{\overset{\overline{}}{l}}{f\left(x_{i}\right)y_{i}-\sum_{i=1}^{\overset{\overline{}}{l}}{f\left(x_{i}\right)\sum_{i=1}^{\overset{\overline{}}{l}}y_{i}}}\right)^{2}}{(\overset{\overline{}}{l}\sum_{i=1}^{\overset{\overline{}}{l}}{{f\left(x_{i}\right)}^{2}-(\sum_{i=1}^{\overset{\overline{}}{l}}{f\left(x_{i})\right)}^{2}})(\overset{\overline{}}{l}\sum_{i=1}^{\overset{\overline{}}{l}}{{y_{i}}^{2}-(\sum_{i=1}^{\overset{\overline{}}{l}}{y_{i})}^{2}})}\)(2)
At the same time, the MLR model was described by equation (3):
Y=b0 + b1×X1 + b2×X2 +…+b 7×X7 (3)