(8) Weak predictor prediction: If the training samples have m setsof data, we initialize the distribution Dt(i) = 1/m, i = 1, 2, . . .,m. Ten MEC-BP neural networks are created by following thisimprovised effort. And the ten BP network weak predictors aretrained. The errors sum etof prediction serial g(t) was obtainedwhen the t weak predictors was trained [14]. Therefore, theprediction serial weight atcan be computed by the followingequation: