Known as the first enclosed earthen site museum in China, the Outer Burial Pits of the Han Dynasty Yang Mausoleum have accumulated more than 7 million pieces of environmental monitoring data of various types. However, due to limitations of big-data mining, utilization of the monitoring data is insufficient and the accuracy of forecasting models is low. In order to solve the problems, a big-data machine learning method, known as the extreme learning machine (ELM), was introduced for temperature analysis and forecasting. The results show that ELM can analyze the annual temperature monitoring data effectively and predicts the instantaneous temperature trends in the future. The introduction of this method represents a useful reference for preventive conservation and management of historical sites.