Competitive dynamics in contemporary markets are characterized by complex interactions, nonlinear changes, and reactive firm behaviors, which make the analysis and prediction of sustainable competitive advantage particularly challenging. Many prior studies rely on static or retrospective models or fail to integrate data driven forecasting with interactive competitive behavior in a unified manner. This study aims to develop a hybrid analytical framework based on deep learning and agent based modeling to capture competitive dynamics and predict sustainable competitive advantage.



Methodologically, this research adopts a quantitative computational design conducted in two sequential stages. In the first stage, historical competitive data including price, market share, and innovation intensity are preprocessed and modeled using LSTM and Transformer architectures. Predictive performance is evaluated through RMSE, MAE, and MAPE metrics. Based on comparative results, the Transformer model is selected as the superior predictor, and its outputs are transferred as decision inputs to the simulation environment.



In the second stage, firms are modeled as autonomous agents whose decision rules are informed by data driven forecasts. Competitive interactions are simulated under three scenarios: moderate competition, intensified competition with innovation waves, and new entrant entry. To evaluate the sustainability of competitive advantage, a composite SCA index is calculated based on market share stability, innovation intensity, and resistance to imitation, while robustness is examined through sensitivity analysis across parameter variations.



The results indicate that sustainable competitive advantage does not arise solely from initial resource positions or isolated innovation activities, but rather from the dynamic interaction of continuous innovation, competitive responsiveness, and reduced susceptibility to imitation over time. The proposed framework advances the literature on dynamic competition by providing an operational and integrated mechanism linking predictive analytics with behavior based simulation. From a managerial perspective, the DL ABM framework functions as a strategic decision support tool that enables managers to test alternative competitive responses before implementation and to anticipate the long term consequences of strategic choices in complex and data intensive market environments. By systematically connecting short term predictions with long term competitive outcomes, this study offers a generalizable approach for analyzing competitive dynamics in intelligent marketing and strategic management contexts.