$\textsf{Turn around time} = \textsf{Waiting time}+\textsf{Burst time}$
Since, $\textsf{Burst time}$ is independent of the scheduling algorthm, to minimize the average turn around time, the OS scheduler has to minimize the average waiting time. This means resorting to $\textsf{Shortest Remaining Time First}$ scheduling which on the positive side gives the minimum average waiting time but on the negative side can cause long burst time processes to $\textsf{starve}.$