This paper proposes a SOH Estimation of LiFePO4 battery management systems using a Linear Regression Analysis. Among the methods of machine learning, supervised learning learns the relationship between the input data (battery characteristic) and the output data (failure data) to find a model that is expressed as a rule or function. Unsupervised learning performs failure diagnosis and prediction by discovering patterns inherent in changing battery characteristics data during use. The algorithm estimates DCIR according to the input parameters using linear regression analysis of supervised learning, and clustering of data to confirm association with failure causes. The validity of the proposed machine learning algorithm is verified by experiment.