Meta-learning is an efficient approach in the field of machine learning, which involves multiple classifiers. In this paper, a meta-learning framework consisting of stacking meta-learning and cascade meta-learning was proposed firstly. Then the algorithm for generating simulated datasets was presented. Finally, based on the classifier simulator, datasets with variable correlation were obtained and used to evaluate the classification performance of meta-learning. Experimental results show that negative correlation measured by Q statistic benefits meta-learning approach.