Facilitating the extraction of extended insights on logistics performance from the logistics performance index dataset: A two-stage methodological framework and its application

Abstract

The Logistics Performance Index (LPI) developed by the World Bank provides a comparative assessment logistics performance in trade logistics for several countries. Given the lack of studies bringing insights on logistics performance in the backdrop of trade logistics from the perspective of nation as a whole—this paper recognizes the LPI dataset as an account of rich country-level data with harbored insights on logistics performance. It further suggests that upon linking the dataset with appropriate variable(s) of interest, extended insights on logistics performance can be extracted. Therefore, a two-stage methodological framework is suggested for the mining of LPI dataset towards extended insights. The first stage involves the clustering of LPI dataset into finite clusters using K-means data mining algorithm. Subsequently, in the second stage, the suitability of multivariate adaptive regression spline (MARS) based regression is outlined for capturing the complex non-linear relationship between the variables under investigation. Thereby, the application of the proposed two-stage methodological framework is demonstrated with an example by linking the six LPI dimensions with an important macroeconomic variable. In addition to discussing the critical implications from the example towards extended insights on logistics performance, and its further implications towards the utility of the proposed methodological framework—the findings suggest that the extraction of extended insights from the LPI dataset is governed by the selection of appropriate variable(s) which can be linked with LPI dimensions, criteria of clustering in the first stage, and development of MARS models for the clusters and the overall data.