Abstract: Machine learning (ML) has been widely used in pile foundation engineering modeling. However, predicting pile tip resistance of long piles remains a significant challenge, as pile tip resistance becomes substantial only when the pile is loaded and displaced to a certain extent. To address this issue and accurately predict pile tip resistance of long piles, three key factors—mechanical effects, pile properties, and soil properties—are identified. A novel hybrid modeling framework is proposed, combining multifold cross-validation, chaotic sequences, nutcracker optimization algorithm, sparse bayesian algorithm, and maximal information coefficient testing. This integrated approach not only improves prediction accuracy but also enhances model interpretability. The dataset, comprising 920 long and super-long piles, was collected in Ho Chi Minh City, Vietnam, and serves as the benchmark for this study. Model performance is evaluated using root mean square error (RMSE), mean absolute error (MAE), and correlation coefficient (R). The results demonstrate that the proposed model outperforms existing ML models in point prediction accuracy, with multiple evaluation metrics approaching optimal values. Furthermore, the paper calculates correlation strengths of various influencing factors on pile tip resistance, thereby significantly enhancing the interpretability of the model's internal calculations in combination with practical engineering insights. This study has significance for the design and research of long piles in soft ground conditions.