Missing pattern aware elastic modeling of sparse village data with dynamic evolution simulation

Traditional sparse data processing methods fail to effectively consider the social, environmental, and economic background differences unique to remote areas, and ignore the data missingness driven by specific socioeconomic factors, resulting in poor model performance. This paper proposes a sparse data elastic relationship model: first, the model weights are dynamically adjusted based on data scarcity and missing patterns. Then, by identifying and analyzing the missing patterns in village data and combining social, economic, and environmental factors, the model identifies the data characteristics of random and non-random missing and uses a cross-domain data fusion method to integrate multi-source data such as remote sensing, social networks, and climate. GAT (Graph Attention Network) is used to dynamically calculate the weights of multi-source data, and GCN (Graph Convolutional Network) is used to propagate features to enhance semantic associations and achieve refined fusion of cross-domain data. Finally, a village dynamic evolution model is constructed based on the Markov decision process. A dynamic simulation model is established, and the ability of the model to respond to different policy and environmental changes is optimized through reinforcement learning. Experiments show that the information retention rate (IRR) of the proposed model in remote sensing data reaches 0.892, which is significantly higher than 0.637 of KNN (K-Nearest Neighbors). The structural similarity index (SSIM) increases to 0.90 with the fusion weight. The results verify the robustness of multi-source data fusion and dynamic weight adjustment under complex missing patterns.