Raman spectroscopy is widely used in the quantitative analysis of components because of its advantages of fast, simple and nondestructive. Currently, quantitative analysis methods of Raman spectroscopy include Partial Least Squares, Artificial Neural Network, Support Vector Machine, etc. In order to seek new methods, in this paper, the Raman spectroscopy data of 41 groups glucose samples were studied. The Extreme Learning Machine was used for quantitative regression. The optimization algorithms such as Genetic Algorithm, Particle Swarm Optimization Algorithm and Artificial Bee Colony Algorithm were used to improve it. After comparison and analysis, a new type of model was proposed, which called Self Adaption Differential Evolution Artificial Bee Colony Algorithm applied to the Extreme Learning Machine. The model adjusted the mutation rate and crossover rate of differential evolution, which can reduce the influence of the Extreme Learning Machine on local optimization and the differential evolution on parameter dependence. Comparing with the traditional Extreme Learning Machine and other optimization algorithm models, the optimized model evaluation index had a significant boost. Experiment showed that Extreme Learning Machine based on Self Adaption Differential Evolution Artificial Bee Colony Algorithm improved the prediction accuracy and model robustness.