Abstract: Big data in the cloud service environment has multi-label attributes, which leads to poor classification in the process of data access. In order to reduce the misclassification rate of big data classification, a classifying method for cloud service big data based on the neural network optimization of extreme learning machine is proposed. The high-dimensional phase space reconstruction model is used to rearrange the information of the random sampling cloud service big data's prior feature, and the association rule mining method is used to extract the label attribute feature quantity of the data. The principal component analysis method of multidimensional label attribute is used to analyze big data's principal feature. The principal feature quantity is input into the neural network classifier, and the weighted control of neural network classification is carried out by combining with the extreme learning machine. Improve the self-adaptability of data classification, realize cloud service big data optimization classification. The simulation results show that the proposed algorithm has higher accuracy and lower misclassification rate in cloud service big data classification, and improves the recall ability of data in cloud composition service.