Data cleaning involves handling missing values, removing duplicates, fixing inconsistent formats, treating outliers, correcting errors, and ensuring data is properly structured before training the ML model.