Determining criteria weights with genetic algorithms for multi-criteria decision making methods: The case of logistics performance index rankings of European Union countries

Abstract

One of the most critical metrics for evaluating countries’ logistics performance is the Logistics Performance Index (LPI). Although the LPI is an effective tool, it is also an index with some concerns regarding performance evaluation. The two main concerns with LPI are assuming equal criterion weights and ignoring operational logistics performance and macroeconomic indicators. To overcome these problems, this study presents a logistics performance evaluation model in which the criteria weights are determined by GA. This study used 11 techniques to determine the logistics performances of the EU countries across 33 indicators. Employing more than one MCDM techniques enabled robustness and consistency. Median, Mean, Linear and K nearest neighbor techniques are implemented to impute the missing values in the dataset. The Linear Regression method is found as the best-performing imputation technique among other techniques for the data of this research. The performance of the GA is compared with other well-known criteria weight determination procedures namely CRITIC, Entropy and equal weight. The ranking results of the MCDM tools were integrated with the Copeland method. The genetic algorithm assigned the highest three weights to the Goods, Value of Exports, Quality of roads, and GDP per capita, indicating that these criteria play an essential role in the countries' logistics performance. The importance of these criteria is not captured by other weight sets. The correlation with the actual LPI scores is higher with Genetic algorithm compared with other weight determination techniques (CRITIC, Entropy, equal weight). Moreover, random weights are created and the results are examined. Weight simulation indicates that, some of the countries ranking are higher than the other countries, regardless of the weights of the criteria. The paper's novelty is to use the Genetic Algorithm (GA) as a criteria weight determination tool for MCDM methods. The proposed system can be used as a decision support system for evaluating the logistics performances of countries. The LPI index is published every two years, but if an automatic or semi-automatic system can be developed using the proposed approach, the logistics performance ranking of countries can be examined more frequently. This can help countries that want to rank high on the LPI ranking determine which criteria to focus on to improve their logistics performance.