Combining both fine-tuning and model stacking has proven a powerful approach, especially for complex problems or datasets where the two approaches can augment each other's performance. So, this is a detailed study report for combining both methods for predictive analysis. One of the important stage of fine-tuning pre-trained models is to select a suitable architecture for the task domain, and adjust the model's layers to adapt to the data at hand. Then, model stacking is just training several models with different architectures and making predictions through a meta-model which uses their collective intelligence to improve performance. When combined with the technique of stacking models, fine-tuning further enhances predictive power and generalizes much better. The amalgamation of fine-tuning, to leverage pre-trained features coupled with model stacking, to apprehend variety in data representations, make this composite work delivers accurate and robust predictives, significantly bridging the gap of machine learning application frontier. The QDR also demonstrates the effectiveness of model stacking, which combines the output of layered models to boost predictions based on the strengths of different model types (Feng et al. in IEEE Trans Neural Netw Learn Syst 35:144–153, 2024; Li et al. in IEEE Access 11:13,479–13,490, 2023).