An attribute value filling method in attributed networks with uncertain information

In real attributed networks, attribute uncertainty and missing attribute values are prevalent, significantly impacting network performance analysis. Existing methods for filling in missing values primarily addresses missing data in the dataset, often neglecting the attribute characteristics and data uncertainty of the network. This study focuses on an effective method to fill uncertain attribute values in attributed networks. First, a framework is proposed to fill missing values in attributed networks. Then, a community partitioning method based on both the attributes and structure of the network is introduced. Additionally, an approach for filling uncertain data in incomplete fuzzy soft sets is presented, providing a solution to the high complexity challenges found in existing methods. Finally, these two methods are combined to propose a new method for filling attribute values in attributed networks. To evaluate the effectiveness of the proposed method, tests are conducted using a synthetic network and real networks of different sizes. The results demonstrate that the proposed method performs excels in terms of filling accuracy, runtime, and stability.