The advancement in the application of light alloys such as magnesium and titanium is closely related to the state of the art of joining them. As a new type of solid-phase welding, ultrasonic spot welding is an effective way to achieve joints of high strength. In this paper, ultrasonic welding was carried out on magnesium–titanium dissimilar alloys to investigate the influences of welding parameters on joint strength. The analysis of variance was adopted to study the weight of each welding parameter and their interactions. The artificial neural network (ANN) was used to predict joint strength. Results show that in ultrasonic welding of magnesium and titanium alloys, clamping force is the most significant factor, followed by vibration time and vibration amplitude; the interactions between vibration time and vibration amplitude, and between vibration amplitude and clamping force also significantly impact the strength. By using the artificial neural network, test data were trained to obtain a high precision network, which was used to predict the variations of joint strength under different parameters. The analytical model predicts that with the increase in vibration time, the increase in optimal joint strength is limited, but the range of welding parameters to obtain a higher joint strength increases significantly; the minimum joint strength increases as well; and the optimal vibration amplitude expands gradually and reaches the maximum when the vibration time is 1000 ms, then shifts toward the low end gradually.