Neural networks(NN) and genetic algorithm(GA) are both theories for obtaining basic function of intelligent information by simulating a model of life which focus on information processing. NN and GA can go through different paths to arrive at a similar goal. Therefore, the two methods can be combined to make up each other's deficiencies. So, a new algorithm, different evolution-based DE-LM, was put forward. Databases for UCI were used to prove that this algorithm is capable of converging fast and does not converge at the local nadir.