Performance based stratification and clustering for benchmarking of container terminals

Benchmarking of container terminals is an important issue facing port management. Data envelopment analysis (DEA), which is a multi-factor productivity measurement tool is generally used in assessing the relative efficiency of homogenous units and setting benchmark for inefficient units. Evaluation of container terminals by DEA produces limited set of efficient units which are reference to inefficient units irrespective of their differences in efficiency scores. DEA projects the lowest efficient units to highest efficient units but in reality, the reference set may be very different in size, environment and operating practices. Every container terminal is characterized by some physical values that represent relevant properties of the terminal. DEA, if employed alone, to measure the efficiency and set benchmark for inefficient terminals to improve efficiency may give biased result because all container terminals vary in their capacity. In order to overcome this shortcoming, in this paper, data mining and DEA are fused to provide a diagnostic tool to effectively measure the efficiency of inefficient terminals and prescribe a step-wise projection to reach the frontier in accordance with their maximum capacity and similar input properties which otherwise is not possible with DEA alone.