Quantifying the risk of soil heavy metal sources can identify the main pollution sources. It can provide a scientific basis for reducing the ecological and human health risks of soil heavy metals. Taking the shallow soil in a Pb-Zn mine watershed in northern Guangxi as a research object, ecological and human health risk assessments were conducted using potential ecological risk assessment (RI) and human health risk assessment (HRA), and the source apportionment of soil heavy metals was completed using the absolute principal component-multiple linear regression receptor (APCS-MLR) model and random forest (RF) model. Then, a combined risk assessment model, consisting of RI, HRA, and APCS-MLR, was used to quantify the risk of soil heavy metal sources. The results showed that the contents of Pb, Zn, Cu, and Cd exceeded the environmental screening values for agricultural land with mean values of 342.77, 693.34, 61.27, and 3.08 mg·kg