A technique in data science where multiple machine learning models are connected in a sequence, with the output of one model serving as the input for the next, to improve predictions or analyses.