As an important part of rotating machinery, the degradation trend prediction of rolling bearings is the key to reducing economic losses and ensuring safe production. To improve the accuracy of prediction, a degradation trend prediction method of rolling bearings based on Mahalanobis distance (MD) and long and short term memory network (LSTM) is proposed. Firstly, the original vibration signal is decomposed and reconstructed by ensemble empirical mode decomposition (EEMD), and the feature space in the time domain and frequency domain is constructed. Then, the correlation, monotonicity, and robustness criteria are used to comprehensively screen the features, and the optimal feature subset is obtained. MD is calculated as the performance degradation index. Finally, the LSTM model is used to predict the degradation trend of bearings. The results show that the proposed method can effectively extract features and accurately predict the trend of performance degradation.