In a big data mining optimization algorithm, the classification algorithm plays an important role. At present, there are many popular classification algorithms based on machine learning. Aiming at the problems of existing big data classification algorithms, two improved strategies and implementation methods are proposed in this paper. First of all, before the data training, the orders of magnitude of the original data are normalized to achieve better data pre-processing and classification. Then with the least mean square algorithm and the BP neural network classification algorithm as the basic algorithms, an improved batch learning BP algorithm is designed, based on the rules of batch learning. The experimental results indicate that the improved batch learning BP algorithm can better solve the imbalanced classification problem in big data.