A decision tree is a flowchart-like structure used for classification and regression tasks in machine learning. It splits the data into subsets based on the most significant predictor variables, making decisions at each node until reaching a leaf node that denotes the output value or class label.