Aiming at the problems that the model-driven method is incomplete in the comprehensive evaluation of university education informatization level, and the determination of evaluation value in data-driven method depends on artificial experience and reliability is difficult to guarantee, this paper fuses model-driven and data-driven, a method combining structural equation and LSSVM by optimizing the least squares support vector machine (LSSVM) algorithm is proposed to evaluate the higher education informatization level. Firstly, the structural equation was used to obtain the evaluation value of educational informatization level in colleges and universities. Secondly, LSSVM is used to construct the evaluation model between the indicators and levels of university education informatization, and the parameters of LSSVM are optimized by the salp swarm algorithm improved by elite centroid and opposition-based learning strategy. Finally, the prediction accuracy and computational efficiency of the proposed algorithm are verified by the questionnaire data of university education informatization construction collected by Shaanxi Provincial Education Department. Experiments show that: The model and data jointly driven evaluation method proposed has the smallest average error of 1.13%. It is 1.27%, 3.03% and 7.40% lower than other algorithms, respectively. The maximum sample deviation is only 4.23%, and the running time is only 3.21 s, it is 42.88% and 19.63% shorter than other algorithms, respectively. It has high prediction accuracy and calculation efficiency. It has certain reference for the comprehensive evaluation of educational informationization level in colleges and universities.