Background: Matrix effect of elements is always a hot and difficult problem in energy dispersive X-ray fluorescence(EDXRF). Purpose: A hybrid model based on fruit fly optimization algorithm(FOA) to optimize the least square vector machine(LSSVM) was put forward to calibrate the matrix effect between Ti and Fe in EDXRF. Methods: The X spectrum counting rate of Ti-Fe in five kinds of ore samples was measured using a CIT-3000SMEDXRF analyzer and electrical refrigeration semiconductor detector. In total 80 groups of data

there were 16 groups of data for each kind of ore samples. FOA was utilized to optimize the parameters of LSSVM

and the optimal model was used for predicting 30 groups of samples of Ti

Fe contents. Moreover

the default LSSVM of MATLAB and the hybrid model optimized by particle swarm optimization(PSO) and genetic algorithm(GA) were applied to predict the contents of Ti and Fe. And the results predicted by FOA-LSSVM were compared with those by PSO-LLSVM

GA-LSSVM and the default LSSVM. Results: There are 26 groups with the Ti

Fe contents predicted by FOA-LSSVM bearing a relative error of 1% compared with chemical analysis values

accounting for 86.67% of the total 30 groups; another 4 groups with the predicted values equal to chemical analysis values

accounting for 13.33%. The study showed that FOA-LSSVM possessed a high prediction precision. Conclusion: Comprehensive research indicates that it is feasible to calibrate the matrix effect between Ti and Fe by FOA-LSSVM model

and the FOA-LSSVM model is a preferred method.