Establishing an Interpretability System for Support Vector Regression and Its Application in QSAR of Organophosphorus Insecticide

Abstract

Aiming at the poor interpretability of support vector regression (SVR), a complete set of interpretability system for support vector regression was established based on F-test. The novel interpretability system includes the significance tests of model and single-factor importance, the single-factor effects and sensitivity analysis, the significance test of two factor interactions and so on. The analysis results of ternary dissymmetric organic phosphate insecticide preliminarily indicate that this new interpretability system is reasonable. Meanwhile, the quantitative structureactivity relationship (QSAR) model of insecticide based on support vector regression are superior to reference model in both fit and prediction, which further confirmed the outstanding regression performance of support vector regression.