Abstract: Aiming at the problem that the accurate evaluation of hydraulic roof support status can not be achieved due to heavy noise, multiple parameters, data redundancy, and uneven sample distribution in monitoring data for hydraulic roof support in fully-mechanized working faces, a parameter reduction method based on an improved neighborhood rough set is proposed to enhance the accuracy of support status evaluation. First, a decision information system for parameter reduction of hydraulic roof support status is constructed by analyzing the characterization parameters and decision parameters of support status. Then, to address the issues of uneven sample distribution and noise interference, a sample similarity index is introduced, a neighborhood partitioning mechanism integrated with sample quality evaluation is established, and a parameter reduction model named MNRS based on parameter dependency is proposed. Finally, support status data from 150 sets of hydraulic roof supports at Hecaogou No. 2 Coal Mine in Yan’an City are collected, and the performance of the proposed model is validated by comparing it with the RS model, the NRS model, and the KNNRS parameter reduction model. The results show that the proposed MNRS model reduces 9 key parameters, achieving a parameter reduction ratio of 35.71%, which represents improvements of 21.42%, 21.42%, and 14.28% over the traditional rough set (RS), neighborhood rough set (NRS), and k-nearest neighbor rough set (KNNRS) models, respectively. Furthermore, based on the reduced parameters, the KNN classification accuracy reaches 97.14% (compared to 92.88% on the original data), the mean squared error (S_ME) is reduced to 0.0286 (a decrease of 49.74%), and the symmetric mean absolute percentage error (E_SMAP) is only 1.14%. For sample sizes ranging from 50 to 150, the MNRS model achieves an average classification accuracy of 96.77%, with the average MSE and average SMAPE of only 0.0367 and 1.38%, respectively, demonstrating significantly superior robustness compared to the benchmark models. Therefore, the proposed method effectively addresses the sensitivity to noise, uneven sample distribution, and incomplete redundancy elimination of traditional models, enabling accurate selection of key parameters for hydraulic roof support status evaluation.