Abstract: Molybdenum-based alloys, valued for their high-temperature strength and creep resistance, represent key candidate structural materials for core components in next-generation nuclear fission and fusion reactors. However, current approaches to designing these alloys remain largely empirical, focusing on compositional adjustments guided by trial-and-error. A critical gap persists: the lack of systematic, quantitative understanding of the synergistic mechanisms between key alloying elements and varying irradiation conditions, particularly irradiation dose. Addressing this predictive limitation is paramount for the accelerated development of high-performance, radiation-resistant alloys. To bridge this gap, this study leveraged a comprehensive dataset comprising 254 samples covering different alloy compositions, irradiation doses, and post-irradiation tensile properties. Recognizing the inherent complexity and potential multicollinearity among numerous high-dimensional features (alloying elements and irradiation parameters), principal component analysis (PCA) was strategically employed. This dimensionality reduction technique effectively extracted the most significant information while mitigating redundancy and collinearity among input variables. The core predictive modeling innovation involved developing a deep learning framework based on convolutional neural networks (CNNs). Crucially, the CNN was trained using a novel composite loss function meticulously designed to balance performance objectives. This function integrated both mean squared error (MSE) and a measure of goodness-of-fit (R²). This synergistic loss formulation enabled effective multi-objective optimization during the training phase, simultaneously targeting high accuracy in predicting both UTS and EL. The resultant CNN model demonstrated high predictive power for the complex irradiation response of molybdenum-based alloys, achieving a remarkable coefficient of determination (R²＞0.8) on key performance metrics. This robust performance validates the effectiveness of the proposed PCA-CNN framework and underscores the potential of data-driven approaches for predicting complex material properties under extreme environments. Analysis reveales that Re significantly enhances post-irradiation tensile strength, while Ti and Zr improve strength but slightly reduce elongation. This provides a new paradigm for compositional optimization in radiation-resistant molybdenum alloy design. This research establishes a high-fidelity, data-driven paradigm for predicting and optimizing the irradiation performance of complex multi-component alloys. Moving beyond empirical composition tuning, provides a powerful tool to systematically map the intricate relationships between composition, irradiation dose, and critical mechanical properties. Overall, this work paves the way for accelerated, targeted composition design strategies to achieve superior radiation resistance in molybdenum-based alloys for demanding nuclear applications.