Abstract: Objectives: Fast fusion of multi-source heterogeneous data of real estate is a difficult problem faced by data integration and database construction in the process of unified real estate registration. The existing multi-source heterogeneous data fusion methods of real estate are unable to accurately construct the relationship describing the spatial entities, and the time cost is too high for the real estate data with huge data scale. Methods: By introducing feature similarity technology, this paper proposes a multi-level fast fusion model to realize batch fusion of real estate multi-source heterogeneous data. Firstly, on the basis of electronization and vectorization of real estate multi-source heterogeneous data, similarity factor is introduced to evaluate real estate multi-source heterogeneous data. Then, a network model based on comprehensive similarity weighting algorithm was designed to calculate the comprehensive similarity of real estate multi-source heterogeneous data in the distribution direction of each similarity factor. Finally, similarity threshold parameters and limited range parameters are used to further improve the accuracy and efficiency of multi-source heterogeneous data fusion. Results: This paper takes real estate data as an example to conduct quantitative analysis of the fast fusion model. The experimental results show that, compared with other methods, the proposed model consumes less time and costs, the average fusion time per thousand houses is 3.29s, and the fusion accuracy reaches 93.5%, which can effectively improve the fusion accuracy and efficiency of multi-source heterogeneous data of real estate.