The loading of non-unit load device in aircraft holds is crucial to air cargo transportation， and ensuring efficient loading of such cargo remains an important and urgent problem. Among the theoretical approaches to address this problem， the two-dimensional rectangular cutting stock problem serves as a key methodology. Although the two-dimensional rectangular cutting stock theory is widely applied to problems such as raw material cutting and packing， the lack of efficient solving algorithms hinders practical implementation， as their computational speed often impedes the whole production procedure. Therefore， this paper proposes a mixed-integer linear programming （MILP） model for the two-dimensional cutting stock problem， aiming to maximize both the utilization rate of the rectangular plane area and the value of the cutting stock. The model incorporates constraints such as boundary adherence， non-overlapping， and orthogonal rotation. A heuristic grouping strategy-based solving algorithm is designed. First， heuristic methods are used to partition the rectangular items into smaller subsets， thereby reducing the variable scale and computational complexity. Second， an exact MILP algorithm is employed to solve the cutting stock problem for each subset. Using classic Benchmark experimental data， the proposed Gurobi grouping method is compared with Gurobi， CutLogic2D， and a hybrid algorithm combining genetic algorithms and the lowest horizontal line algorithm. The experimental results show that CutLogic2D performs well in terms of both solution quality and speed. The Gurobi grouping method， as a heuristic algorithm， generally shows slightly inferior proformance than CutLogic2D. In contrast， the genetic algorithm and the lowest horizontal line algorithm， as the heuristic methods without grouping strategies， along with Gurobi， exhibit relatively long solving times in some cases， reaching up to 7 200 s， which is unacceptable for practical applications.