Liver cancer is one of the most deadly malignancies in the world with little survival opportunities. The conventional mathematical models of statistics do not have the ability to indicate intricate non-linear relationships between various forms of patient data. In this paper, a machine learning (Gradient Boosting Decision Trees) and deep neural networks (DNNs) algorithm with explainable AI (XAI) elements proposed are integrated to create a predictive analytics framework that precision predicts the likelihood of liver cancer. The model digests multimodal data such as clinical, biochemical and imaging features with state of art preprocess, feature selection and explainability piping like SHAP and LIME. On TCGA-LIHC data, the experimental assessment recorded an accuracy of 92.5 percent and C-index of 0.89, which is higher than that of the conventional models. The framework suggested offers clear patient-centered predictions, which will consequently lead to the achievement of precision oncology and AI-based clinical decision support.