Abstract: In order to solve out the knowledge reduction task for the explosive increment big data, a novel attribute co-reduction algorithm (CMQEACR) based on cloud computing and multi-layer quantum elitists was proposed in this paper. First, the large-scale dataset was decomposed into different evolutionary frog subpopulations under the MapReduce cloud mechanism, and the optimal solutions of subpopulations were attained, respectively. Second, the strategy of attribute co-reduction based on the elitist vectors of multi-layer quantum frogs was constructed, and the vectors of elitist subpopulations with the strongest optimization ability of both global searching and local exploration ware selected out. This strategy could guide each subpopulation to obtain its respective minimum attribute reduction. So the global optimal reduction set could be achieved efficiently. The experimental results indicated the effectiveness and accuracy of proposed algorithm for attribute reduction on big data, compared with the representative methods. Meanwhile, the proposed algorithm was used for the segmentation of MRI in the electronic medical record database and the promising results showed its better applicability.