Abstract: Alloy 690 is influenced by factors like temperature, atmospheric pressure, and the transport medium during operation. Stress corrosion in alloy pipelines often exhibits heteroscedasticity and instability, complicating the accurate prediction of the stress corrosion rate for alloy 690 pipelines. In order to improve the prediction accuracy of stress corrosion crack growth rate of Alloy 690 under high temperature environment, the NRS-TPE-LGBM prediction model was established by combining neighborhood rough set (NRS), Bayesian optimization(TPE) and lightweight gradient boosting learning machine (LGBM). The attributes of NRS were used to reduce the dimension of data and improve the operation efficiency of the model. Then, TPE was used to optimize the hyperparameters, so that the gradient boosting capability of LGBM can reach the optimal, and the relative error evaluation index was introduced to establish the comparison between the data set and the prediction set. The results show that compared with the unoptimized LGBM model and the LGBM model optimized by random search algorithm, the prediction accuracy of NRS-TPE-LGBM model is improved by 4.81% and 2.52%, respectively, which verifies the accuracy of the model and provides a more accurate judgment basis for the actual engineering maintenance and replacement of alloy pipelines.