A novel fuzzy knowledge graph structure for decision making of multimodal big data

A novel fuzzy knowledge graph structure for decision making of multimodal big data

연구 분야: Databases

학회: Applied Intelligence

Decision-making in the era of big data is always a challenge. Recently, various methods especially graph sampling have been presented to assist the decision more effectively. As real-world graphs are large, constantly evolving, and distributed in nature, it becomes necessary to sample their structures for many different goals. Therefore, acquiring a comprehensive and in-depth understanding of graph sampling is essential to strengthen this field. In addition, graph sampling techniques often rely on edge or vertex sampling without effective methods for rule or path sampling. In this paper, we propose a novel framework for the rule-based sampling method on fuzzy knowledge graphs. In this framework, fuzzy knowledge graphs are built on integrated databases from multiple sources. We design a purposive random sampling method based on fuzzy rules on graphs to prioritize important rules for output inference. The remaining important rules form the core structure of the fuzzy knowledge graph, known as the Fuzzy Knowledge Graph Structure (FKGS). This structure is considered as a compression mechanism to reduce computational complexity when representing and performing calculations for large-scale data problems. Experimental results based on benchmark datasets on diabetes mellitus show that the sampling method greatly reduces the calculation time while maintaining high accuracy. Moreover, the purposive random sampling method results in significantly higher accuracy than the random sampling method. Besides, the ANOVA method is also conducted to statistically validate the model. The results are significant for decision-making in the context of big data.