This paper discusses the new problems faced by multi-type data fusion, discusses the relationship available for fusion in multi-type data from the analysis of correlation, puts forward the concept of data type attribute correlation, and summarizes the research status of data fusion based on correlation. The general process of multi-type data fusion research is given, and the relevant literature is summarized according to the process. The model skillfully combines the hierarchical structure of wireless sensor networks and neural networks, and designs each cluster as a three-layer perceptron neural network model. The feature data is extracted from the large amount of raw data collected by the neural network method, and then the feature data is sent to the aggregation node.