A delivery dispatcher needs a duration before assigning a route. A subscription team needs a yes-or-no departure prediction. Both use past observations, but the outputs are different. Start with the decision that a person will make, then name the observation and the outcome. A row might be an order, a customer-month, or an appointment.
Regression predicts a numeric quantity. Classification predicts a category, even when its labels are written as 0 and 1. Clustering groups similar observations without a target label. You can't choose among them by looking only at the number of columns. These exercises use fictional operational records. Count rows carefully, and keep repeated observations when the task defines each row as a separate event.