Stunting remains a significant global public health challenge, affecting approximately 148 million children under the age of five. This condition leads to long-term cognitive and physical deficits, particularly in low- and middle-income countries. Many existing prediction models fail to capture the complex interdependencies between nutritional, socio-economic, and environmental factors. To address this gap, our study introduces a 1D-Convolutional Neural Network (1D-CNN) model designed to predict childhood stunting using structured datasets collected from community health centers (Puskesmas) and validated by the Cirebon City Health Department (Dinas Kesehatan Kota Cirebon), Indonesia. The dataset includes anonymized records of children under five years old, comprising anthropometric measurements, socio-economic profiles, nutritional intake, and environmental indicators, gathered through household surveys and routine public health reporting. The proposed 1D-CNN architecture is optimized for structured data by integrating convolutional and pooling layers, dropout regularization, and dense classification layers. To enhance interpretability, we employ explainable AI (XAI) methods—SHAP and LIME—to reveal the relative influence of each feature in the model’s decision-making process. Additionally, the study applies a participatory validation approach through focus group discussions (FGDs) with community health workers, ensuring contextual relevance and ethical integrity. Experimental results demonstrate the superior performance of the proposed model, achieving 93.12% accuracy, with a precision of 97% and a recall of 89%, resulting in an F1-score of 93% across both stunted and non-stunted classes. These findings outperform traditional machine learning approaches and highlight the potential of AI-driven predictive frameworks for early stunting detection and policy-oriented health interventions. This research contributes to the advancement of data-driven public health strategies by integrating predictive analytics, community participation, and transparent AI methodologies