On the change-point problem