An ultraviolet metamaterial absorber is designed in this paper, consisting of a titanium nitride (TiN) substrate, a silicon dioxide (SiO₂) dielectric layer, and a TiN patterned layer. The absorption characteristics of the absorber are analyzed using the finite element method (FEM), and a dataset is established. A deep learning model based on a tandem neural network is employed for the inverse optimization design of the absorber, significantly accelerating the design process. It is demonstrated that with the optimized structural parameters, an absorptivity greater than 94.3% is achieved in the 200?400 nm range under normal incidence, with absorptivity of 97.0% in the 265?375 nm range. Under a 50° angle of incidence, an average absorptivity of 82.6% is obtained for transverse magnetic (TM) wave incidence, while an average absorptivity of 83.0% is observed for transverse electric (TE) wave incidence. The absorption mechanism of the absorber is attributed to cavity resonance effects. When compared to absorbers reported in existing literature, the ultraviolet metamaterial absorber proposed in this study is not only simpler in structure and easier to fabricate but also exhibits higher absorptivity in the ultraviolet band. Additionally, it is characterized by stability at large incident angles and polarization independence.