Although causal explanation should be the foundation of prediction, quantitative research in the social sciences has long been limited by data and computing power. As a result, most studies have focused on testing correlations and identifying causal relationships, while prediction has received little attention. This chapter reviews the historical development of social prediction and redefines the concept by explaining how machine learning makes it possible to conduct prediction in a scientific way. In practice, supervised machine learning provides an important approach to realizing social prediction by training models on known outcomes and applying them to new data to forecast social behavior and change. Through machine learning, an era of paradigm shift from correlation and causality to social prediction is emerging.