Dynamic Weighted Multidimensional Similarity Matrix Data Grafting Algorithm

Data source matching and missing value imputation are crucial challenges in the integration of multisource heterogeneous data in the era of big data. This paper proposes a novel data grafting algorithm based on dynamic weighted similarity matrices. The algorithm constructs dynamic weight similarity matrices that capture the complex relationships between multidimensional features, optimizes similarity function parameters, and comprehensively integrates feature similarities to guide the selection of the most similar samples for data source matching and precise imputation of missing values. Experimental results on a dataset of Chinese city statistics demonstrate that the proposed method significantly outperforms traditional approaches such as mean imputation and K-Nearest Neighbors (KNN) in terms of Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE), validating the effectiveness and superiority of the algorithm. The dynamic weighted similarity matrix model presented in this paper provides a new perspective for high-quality fusion of multi-source heterogeneous data.