To discuss how to provide a reliable platform for processing big data, by studying some sub projects in core technology and its ecological environment, an extensible big data application system architecture used for processing large scale data sets is implemented. In addition, the architecture is organized into four levels in the form of hierarchical model: platform layer, data layer, business logic layer, and application layer. From the angle of hierarchy, the optimization of Hadoop and HBase, MapReduce programming mode, Mahout technology details and clustering algorithm are discussed. In combination with the customer value hierarchy model and neural network classification model, a distributed brand value model is proposed and successfully deployed on the framework application layer. As a result, it improves the research and implementation of the whole system architecture. Finally, the research contents of each layer are summarized, and the future research directions of the corresponding parts are discussed.