This research aims to build a decision support system for real estate bubble early warning in accordance with the forecasted bubble degree. Both theoretical review and empirical evidence will be conducted to determine the ultimate input variables for the forecasting model. It begins with the identification of the symptom indicators of real estate bubble. The quantitative and qualitative information will be analyzed in uncertain and imprecise environments. Then the principal component analysis (PCA) will be applied to categorize these key indicators. Based on the key variables, advanced mathematical methods like artificial neural network (ANN) and support vector machine (SVM) will be deployed for mapping the complicated non-linear relationship between the bubble indicators and the bubble degree. Meanwhile, the forecasting ability of ANN and SVM will be compared, so as to determine a more desirable one for bubble forecasting. Market data from Hong Kong and Mainland China are employed for the empirical study. The ultimate aim is to build a real estate price bubble early warning system (EWS), and provide an objectives basis of planning, decision-making for the government and real estate management departments respectively.