Based on the research of prediction methods for foundation or subgrade settlement by existing combination model，firstly，an analytical model of freshness function is proposed to measure the newness-oldness degree of measured settlement data，considering the features of newness-oldness degree and its influence boundedness of measured settlement data. The proposed method can also reflect the effect of time distance between the measured and predicted point，measured error and boundedness of newness-oldness as well as its influence degree of the measured data on the settlement prediction. Secondly，based on statistical theory，a disposal method of measured data sample is developed which can exclude the unreasonable settlement prediction caused by some abnormal measured data. Then，using the above-mentioned model and method，a new method for combined prediction of foundation or subgrade settlement is put forward，which can reflect the influence of abnormality of measured settlement data，their oldness and newness degree and boundedness of its influence degree on settlement prediction of foundation or subgrade. Finally，it shows the rationality and superiority of the proposed model and method by calculation of actual engineering examples and comparison and analysis among the prediction results by the proposed and the existing similar methods as well as the measured results.