A Comparison of the Imputation Methods for Balanced Transportation Problems

Transportation model is a special case of linear programming to transport a product from sources to destinations at total minimum cost. The objective of this study is comparing the efficiency of estimating when the variable costs are missing completely at random in balanced transportation problems by imputation 4 methods which are mean, median, maximum, and minimum. The cost for transport model is set to a uniform distribution and a normal distribution with coefficients of variation to 5%, 10%, 15%, 20%, 25%, and 30%. The Monte Carlo simulation was repeated 1,000 times. Set the number of sources and destinations to be equal in size 4, 5, 7, 10, and 20. The percentage of missing data are 5%, 10%, 15% and 20% of the sample size. The criterion used for comparison is the Root Mean Square Error (RMSE). The results found that as the number of sources and destinations increases, RMSE also increases across all imputation methods. Among the methods, imputation using the mean resulted in the lowest RMSE, followed by the median, maximum, and minimum, respectively.