Abstract: In order to improve the prediction accuracy of the traditional unequal time-interval gray model (TUTGM), an improved unequal time-interval weight gray model and residual combined correction (IUTWGM-RCC) is proposed. Firstly, the weight distribution coefficient of time interval is introduced into TUTGM. According to the different generation sequence of cumulative and reduction process, four different forecasting models are constructed, and the optimal fitting sequence and prediction value are determined based on similarity criterion. Secondly, the periodic sequence function generated by sine function and harmonic generation is used to correct the residual sequence, which further improves the prediction accuracy of the model. Finally, the settlement of three measuring points of building is predicted. The results show that the prediction accuracy of TUTGM is the highest by introducing the weight distribution coefficient in the reduction process among four different forecasting models. After the residual combined correction, the posterior error ratios are 0.04, 0.11 and 0.05 respectively, and the precision rank is grade 1.

Key words: settlement prediction, grey theory, unequal interval, improved weight, residual correction, similarity criterion