On 12 May 2008, a devastating earthquake occurred in Sichuan Province, China, taking tens of thousands of lives and destroying the homes of millions of people. Among the large number of dead or missing were children, particularly children aged less than five years old, a fact which drew significant media attention. To obtain relevant information specifically to aid further studies and future preventative measures, a neural network model was proposed to explore some geological and socio-demographic factors associated with earthquake-related child mortality. Sensitivity analysis showed that topographic slope (mean 35.76%), geomorphology (mean 24.18%), earthquake intensity (mean 13.68%), and average income (mean 11%) had great contributions to child mortality. These findings could provide some clues to researchers for further studies and to policy makers in deciding how and where preventive measures and corresponding policies should be implemented in the reconstruction of communities.