Shen Hong

With the rapid development of Large Language Models (LLMs), the scarcity of high-quality Supervised Fine-Tuning (SFT) data emerges as a critical bottleneck restricting the improvement of model performance. Existing automated data synthesis methods face challenges such as domain semantic fragmentation, severe instruction homogeneity, and lack of interpretability in quality evaluation. In intelligent manufacturing and predictive maintenance scenarios, production scheduling, equipment condition monitoring, and maintenance decision-making are highly coupled. Complex mappings exist among sensor time-series data, failure modes, and downtime costs, imposing higher requirements on the cross-modal semantic representation and complex decision reasoning capabilities of instruction data. Hence, this paper proposes a full-cycle intelligent data construction framework, and an end-to-end system named IntelliData-Forge generates high-quality fine-tuning instruction datasets. First, this paper proposes an adaptive chunking algorithm based on structural entropy and semantic coherence for unstructured long documents, which incorporates a reinforcement learning strategy optimizer to maximize the preservation of contextual integrity. Second, this paper introduces an instruction evolution engine that uses a genetic algorithm integrated with MAP-Elites to achieve directional expansion within the complexity and topic spaces through multi-dimensional crossover and mutation operators. Finally, this paper constructs a multi-dimensional quality evaluation system based on Dempster-Shafer (DS) evidence theory, which utilizes the conflict coefficient K to quantify evaluation uncertainty and trigger an active learning mechanism, thereby achieving closed-loop monitoring of data quality. The experimental results demonstrate that IntelliData-Forge outperforms existing baseline methods across multiple key metrics, including RUL prediction error, fault classification accuracy, decision consistency, and instruction diversity. In particular, the RUL prediction error decreases significantly, while fault classification accuracy and decision consistency achieve substantial improvements.

Distributed permutation flowshop scheduling is an NP-hard problem that has become a hot research topic in the fields of optimization and manufacturing in recent years. Multimodal optimization finds multiple global and local optimal solutions of a function. This study proposes a harmony search algorithm with iterative optimization operators to solve the NP-hard problem for multimodal optimization with the objective of makespan minimization. First, the initial solution set is constructed by using a distributed NEH operator. Second, after generating new candidate solutions, efficient iterative optimization operations are applied to optimize these solutions, and the worst solutions in the harmony memory (HM) are replaced. Finally, the solutions that satisfy multimodal optimization of the harmony memory are obtained when the stopping condition of the algorithm is met. The constructed algorithm is compared with three meta-heuristics: the iterative greedy meta-heuristic algorithm with a bounded search strategy, the improved Jaya algorithm, and the novel evolutionary algorithm, on 600 newly generated datasets. The results show that the proposed method outperforms the three compared algorithms and is applicable to solving distributed permutation flowshop scheduling problems in practice.

Multimodal optimization is to find multiple global and local optimal solutions of a function, rather than a single solution. This study proposes a harmony search algorithm with iterative optimizing operators to solve the NP-hard distributed permutation flowshop scheduling for multimodal optimization. First, the initial solution set is constructed by using a distributed NEH operator. Second, after generating new candidate solutions, efficient iterative optimizing operators are applied to optimize these solutions and the worst solutions in the harmony memory(HM) are replaced. The proposed operations are repeated until the stopping condition of the algorithm is met. Finally, the solutions satisfying multimodal optimization in the harmony memory are obtained. The constructed method is compared with two meta-heuristics, the iterative greedy meta-heuristic algorithm with a bounded search strategy and the improved Jaya algorithm, on 600 newly generated datasets. The results show that it runs stably and outperforms the two algorithms compared.

Credit risk transmission between cross‐platforms is an important issue in the construction of a credit service system. The effect of credit risk transmission between credit entities (nodes) is analyzed in this paper. A heuristic algorithm based on hybrid strategies (HAHS) is proposed to find risk transmission paths and calculate the influence of nodes. Besides, a novel community model is applied to predict the credit risk areas in advance. In detail, the mathematical association structure between credit entities is firstly given in the algorithm, and the breadth first search algorithm is used to find the hierarchical nodes on the credit risk transmission paths. Then, the characteristics of credit risk transmission are analyzed, and the calculation methods of single‐path and multipath influence are proposed. Finally, the credit entities are divided into communities based on a greedy strategy considering the characteristics of the credit entity association structure. The threshold control strategy is adopted to find global key nodes among all of the entities and local key nodes in communities, respectively, so as to realize the early warning of credit risk.

In the credit cloud, credit services are sold to applications for credit computing, credit fusion and credit risk estimates. Plenty of services with different performance for the same task may have different execution time and charged by various ways. The users have specific requirements for the workflow completion time or cost. Hence, to meet the user’s satisfaction is an important challenge. In this paper, we propose heuristic scheduling methods for credit workflow with total cost minimization, and the deadline should be satisfied. The problem can be divided into two sub-problems, task-mode mapping and task tabling on renting service instances. For the task-mode mapping problem, a recursive heuristic method is constructed to select appropriate service for each task of the workflow. Then another heuristic algorithm based is established to get a final schema with deadline constraint. We discussed the service instance rented in shareable manner and compared with un-shareable manner. Three renting strategies are discussed in detail. Experimental results show the effectiveness and efficiency of the proposed algorithm.