The selective detection of ammonia, a hazardous atmospheric pollutant, at room temperature remains a key challenge for high-performance gas sensor development. In this study, three ZnIn2S4 (ZIS)-based metal sulfide heterojunctions including Ag2S/ZIS, CuS/ZIS, and NiS/ZIS were synthesized via a cation exchange strategy, where the formation of well-defined heterostructures led to significantly enhanced sensing performance. Among them, the Ag2S/ZIS sensor exhibited ultra-high sensitivity, achieving a response of 2.35 to 10 ppm ammonia at room temperature. Furthermore, a neural network model was employed to analyze the response curves of the three heterojunctions under varying gas concentrations. The results revealed that all three sensors achieved recognition accuracies exceeding 87.5%, with the Ag2S/ZIS heterojunction showing the highest accuracy, highlighting its superior performance compared to the other two sensors.