Abstract: In order to solve the problems of low accuracy of the prediction model of the settlement of the mining area and the inconsistency between the prediction model and the actual mining，a new subsidence prediction model of mining area based on synthetic aperture radar interferometry （InSAR），support vector regression （SVR） and simulated annealing （SA） was proposed.Firstly，the mining subsidence monitoring data were obtained by InSAR technique，the accumulated subsidence data of the test point was obtained by processing the data.The accumulated subsidence data is basically consistent with the actual GPS monitoring results by comparing with them.Then，the subsidence prediction model of mining area was established，the static subsidence prediction model was obtained by SVR algorithm，the optimal parameters of the model was determined by SA algorithm.In order to make the prediction results conform to the actual mining situation，the embedded dimension formula is introduced to obtain the subsidence prediction model of mining area and accuracy evaluation indexes.Finally，the established model is applied to Daliuta mining area of Shaanxi Province，the maximum absolute error between the predicted value and the actual monitored value is 9 mm，and the maximum relative error is 3%， the calculation results of accuracy evaluation indexes of the model show that the maximum value of the average absolute error of the test area is 2.5%， and the minimum correlation index is 0.8，which indicate that the prediction accuracy of the model is high.

Key words: subsidence mining, InSAR, SA-SVR algorithm, subsidence prediction