To improve the efficiency of concept lattice construction, this paper proposes an incremental concept lattice construction method driven by a granular concept network. The dynamic updating mechanism of the granular concept network is investigated, and a cross-level concept fusion strategy is proposed to generate new concept nodes, thereby achieving the incremental expansion of the network structure. On this basis, the concept lattice of an updated formal context is obtained from the granular concept network. The experimental results show the effectiveness of the proposed method in concept acquisition.

In operations research and decision-making problems, the quality and efficiency of decisions depend on the effective acquisition and rational utilization of knowledge. As an effective knowledge acquisition tool, formal concept analysis (FCA) not only provides the conceptual knowledge required for decision-making but also facilitates knowledge visualization through tools like the Hasse diagram (concept lattice), thereby aiding in optimizing the decision process. However, existing concept acquisition algorithms face efficiency issues when dealing with large-scale datasets, lack incremental update mechanisms, leading to redundant computations. Therefore, it is crucial to design and propose an efficient concept acquisition model that addresses these inefficiencies and provides reliable support for knowledge discovery in complex decision tasks.

To maximize algorithm efficiency, this paper explores the incremental acquisition of concept knowledge based on granular computing and incremental learning. Granular computing effectively reduces computational complexity by processing information at a granular level, while incremental learning improves algorithm flexibility and efficiency by dynamically updating the model to adapt to the continuously changing data. Furthermore, drawing upon the structural analogy between concept generation and neural computation, this paper formally defines an incremental granular concept network. This model efficiently acquires concept knowledge from incremental data through a cross-level fusion strategy, laying a theoretical foundation for scalable knowledge representation.

This paper defines the granular concept network from the perspective of attribute granular concepts and investigates its dynamic update mechanism. It proposes a cross-level concept fusion strategy to generate new concept nodes, thus achieving the incremental expansion of the network structure. Furthermore, it introduces the incremental granular concept network (IGraCN), designed for dynamic data environments, to acquire concept knowledge and construct concept lattices for formal contexts with continuously updated attributes.

Through experimental comparison on 10 datasets with 6 benchmark algorithms, it is demonstrated that the proposed incremental concept lattice acquisition model, IGraCN, can effectively extract concept knowledge from datasets, with construction efficiency superior to that of the benchmark algorithms. Additionally, it is expected to become increasingly advantageous as the data scale grows. The incremental learning capability evaluation further validates the model's ability to handle dynamic data environments. The experimental results also indicate that the proposed algorithm is influenced by the complexity of the relationships between objects and attributes. As this complexity increases, the time required to construct the concept lattice also grows, reflecting the computational overhead when handling high-complexity data.