本发明涉及一种硬盘剩余寿命预测方法、系统及产品。该方法包括：读取硬盘数据，并对所述硬盘数据进行预处理，确定训练盘以及测试盘；将训练盘内的训练样本输入至LSTM模型，训练LSTM模型，生成训练后的LSTM模型；将测试盘内的测试样本分别输入至训练后的LSTM模型，预测每个测试盘的剩余寿命；根据所有测试盘的剩余寿命，生成预测剩余寿命曲线；根据训练盘以及测试盘的真实剩余寿命，生成真实剩余寿命曲线；根据预测剩余寿命曲线以及真实剩余寿命曲线确定最佳LSTM模型；获取待测硬盘的待测硬盘数据，并将待测硬盘数据输入至所述最佳LSTM模型，预测待测硬盘的当前剩余寿命。本发明能够准确反应硬盘全局的健康状态，提高硬盘剩余寿命预测精度。

The invention relates to a method, system and product for predicting the remaining life of a hard disk. The method includes reading hard disk data and preprocessing the hard disk data to determine the training disk and test disk; Input the training samples in the training disc into the LSTM model, train the LSTM model, and generate the trained LSTM model; The test samples in the test disc were input into the LSTM model after training to predict the remaining life of each test disc. According to the remaining life of all test disks, the predicted remaining life curve is generated. According to the real residual life of the training disk and the test disk, the real residual life curve is generated. The optimal LSTM model is determined according to the predicted residual life curve and the real residual life curve. Obtain the data of the hard disk to be tested and input the data of the hard disk to the optimal LSTM model to predict the current remaining life of the hard disk to be tested. The invention can accurately reflect the global health state of the hard disk and improve the prediction accuracy of the remaining life of the hard disk.