Piping is one of the phenomena of water erosion that causes significant changes in the earthen embankment, resulting in about 90% of earth dam failures. To analyze this issue, this study addresses the critical issue of piping susceptibility in the Habdat Earthen Embankment context using the deep learning approach. Through a comprehensive approach, encompassing data acquisition, data preprocessing (Data cleaning, standardization and normalization), feature extraction, feature selection, and prediction, a robust model is developed for assessing piping occurrences. Leveraging advanced techniques, including the Self-Improved Green Anaconda Optimization (GAO) and hybrid deep learning models such as Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN), this research achieves enhanced predictive accuracy. Evaluation results indicate the proposed model’s superiority consistently outperforming benchmark models in accuracy (0.98), precision (0.99), sensitivity (0.98), specificity (0.97) F1-Score (0.93), MCC (0.99), NPV (0.95), FPR (0.02) and FNR (0.00). The incorporation of GAO and hybrid deep learning not only enhances predictive accuracy, but also showcases adaptability across diverse scenarios. This research contributes significantly to geotechnical engineering, offering a foundation for further research and practical applications in risk assessment, infrastructure planning, and disaster prevention. The findings underscore the potential of innovative optimization and deep learning techniques in robust piping susceptibility assessment.