Mind evolutionary computation (MEC) is a global algorithm whose architecture is specifically designed to confront the intrinsic flaws in genetic algorithm (GA). Mind evolutionary computation(MEC) simulates the process of human mind evolution, and come sup with “similar-taxis” and “dissimilation” to instead of “crossover”and “reproduction” in GA. Moreover, MEC designs billboards to record the evolutionary information that will in turn guide the evolution. Utilizing mind evolutionary computation (MEC) to search for the initial weights and thresholds of the BP neural net-work could guarantee a relatively high probability to obtain the global optima. Then the initial search by mind evolutionary computation is a preferred means to overcome the disadvantages of BP neural network. Although setting the initial weights and thresholds of the MEC-BP neural network randomly can avoid getting local optimal and increase the probability to find the neural network, the randomness will result in the results may be not unique even based on the same training samples. Adaboost algorithm is included in MEC-BP neural network, and constructs weak predictors of MEC-BP neural network and trains the sample training data repeatedly to form a new strong predictor. Finally, we further reduce the possibility of our algorithm getting into local minimum and improve the generalization ability. Hence we propose the MEC-BP-Adaboost algorithm.