The serious adulteration of the rubber accelerator 2-mercaptobenzothiazole (MBT) will lead to the failure quality of rubber products. Aiming at this problem, a method which uses terahertz time-domain spectroscopy (THz-TDS) was proposed for quantifying the effective content of MBT. The absorption characteristic spectrum of the mixture of MBT and polyethylene in 0.3THz~1.4THz was obtained by terahertz transmission measurement. A MBT quantitative detection model based on least squares-support vector regression (LS-SVR) was proposed. LS-SVR model was compared with partial least squares model and support vector regression model respectively. The results show that, the root mean square errors of the model prediction sets are 1.1330%, 2.5583% and 2.3869% respectively. The quantitative model of LS-SVR can achieve better results with higher accuracy and better stability. This study provides a new fast and effective method for quantitative detection of MBT.