As the advancement of modern science and technology, the volume of data used is expanding, so the research of big data become extremely urgent. This paper study sequential prediction and clustering of big data. Firstly, we propose a sequential linear regression (SLR) method for a large amount of sequential data. This method is not on-ly computationally efficient in speed and storage but also has higher accuracy than the method of mean predicting. Besides, a weighted strategy is introduced on the curren-t model to determine the impact of data from different periods. Secondly, we propose sparse autoencoder neural network method for reducing dimensions for the high dimen-sions unlabelled data, the solution algorithm is numerical optimization algorithm and a standard k-means algorithm is applied to form the clusters on the hidden layer. When compared with others clustering method, we demonstrate the advantage of our method from the simulation data and real data. |