From assigning the weights of multiple-index correctly, a new method based on minimal deviations and maximal jointentropy is proposed, in which the weight vectors at different point of views determined by several evaluation method was considered and the whole-warp between the ideal combination of the estimated weight vectors and the combination determined by other methods became minimal.   Other hand, the joint-entropy at the multi-indicator vectors the most by choosing the balancing element and the uncertainty at the overall situation was the least in this method.  Thus, the optimization model about the weight vectors combination was constructed. An example is also given to demonstrate the precision and reliability and validity of the method.