Title: On the benchmarking of port performance. A cosine similarity approach

Authors: Iñigo L. Ansorena

Addresses: Polytechnic University of Madrid, ETSI Caminos, Canales y Puertos, C/Profesor Aranguren s/n, Madrid 28040, Spain

Abstract: The measurement of port performance is not a simple task since there are many variables involved in port operations. Some industry standards are based on key performance indicators, which are the basis to provide management information for planning and control. We propose the use of cosine similarity to gain a better understanding of performance indicators. The concept of 'cosine similarity' is particularly interesting when searching for similarities in big datasets. The name derives from the term 'direction cosine', i.e., two vectors are maximally 'similar' if they are parallel (maximum value of cosine), and maximally 'dissimilar' if they are perpendicular (90° angle means uncorrelated). The methodological procedure presented in this paper allows the detection of similarities and the visualisation of big datasets in an easy way. An empirical case study is used to explore the effectiveness of the proposed approach.

Keywords: benchmarking; port performance; past performance; data visualisation; cosine similarity; similarity search; big dataset; containerships; measuring; weighted edges; graph theory; complex network.

DOI: 10.1504/IJPMB.2021.112258

International Journal of Process Management and Benchmarking, 2021 Vol.11 No.1, pp.101 - 114

Received: 11 Aug 2018
Accepted: 07 Nov 2018
Published online: 06 Jan 2021 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article