Abstract: In order to solve the problem of land destruction caused by mining subsidence and to maintain the ecological environment stability of the mining area, the study on the suitability of land reclamation in subsidence area was carried out based on the theory of R/S analysis and limit learning machine. Firstly, the stability status of subsidence area was judged by R/S analysis, and the evolution law of subsidence stability with time was analyzed. Then, the prediction model of subsidence deformation was built by using limit learning machine to evaluate the development trend of subsidence deformation, so as to provide basis for land reclamation. Finally, the study on land reclamation measures in the subsidence area was carried out combined with the characteristics of coal mining subsidence. The case study showed that the R/S analysis could effectively evaluate the subsidence stability of the subsidence area, and that there were some differences in the subsidence stability of different stages, but the stability tended to increase with time. At the same time, through the prediction of subsidence deformation, the relative error mean value of the prediction results of the prediction model was less than 3%, which has high prediction accuracy, and the prediction model was verified. The subsidence deformation tended to be stable in the later period, and the stability evaluation was integrated to obtain the appropriate land reclamation measures to be taken in the future. Finally, the land reclamation in the subsidence area is a systematic work, and it is imperative to grasp the reasonable reclamation opportunity through the settlement stability evaluation, and reasonably standardize the reclamation measures in combination with the reclamation conditions of the mining area, to improve land use efficiency.

Key words: mining subsidence area, R/S analysis, limit learning machine, stability evaluation, land reclamation