Purpose The purpose of this paper is to identify corrosion types and corrosion transitions by a novel electrochemical noise analysis method based on Adaboost. Design/methodology/approach The corrosion behavior of Q235 steel was investigated in typical passivation, uniform corrosion and pitting solution by electrochemical noise. Nine feature parameters were extracted from the electrochemical noise data based on statistical analysis and shot noise theory. The feature parameters were analysis by Adaboost to train model and identify corrosion types. The trained Adaboost model was used to identify corrosion type transitions. Findings Adaboost algorithm can accurately identify the corrosion type, and the accuracy rate is 99.25%. The identification results of Adaboost for the corrosion type are consistent with corroded morphology analysis. Compared with other machine learning, Adaboost can identify corrosion types more accurately. For corrosion type transition, Adaboost can effectively identify the transition from passivation to uniform corrosion and from passivation to pitting corrosion consistent with corroded morphology analysis. Originality/value Adaboost is a suitable method for prediction of corrosion type and transitions. Adaboost can establish the classification model of metal corrosion, which can more conveniently and accurately explore the corrosion types. Adaboost provides important reference for corrosion prediction and protection.