Enhancing Kendall's W using genetic algorithm: a computational approach to inter-rater reliability optimization

This paper presents a computational approach to enhancing Kendall's coefficient of concordance (Kendall's W) using genetic algorithm (GA). Kendall's W is a non-parametric statistic used to measure the degree of agreement among raters. Traditional methods of calculating Kendall's W may not adequately address issues such as outliers and tied rankings. This study proposes the use of GA to optimize the inclusion of raters and improve the robustness of Kendall's W. A custom web-based tool was developed to implement the GA-based optimization model, allowing users to select parameters such as alpha (α), crossover, mutation, and the number of replications. The results demonstrate significant improvements in agreement among raters, highlighting the effectiveness of GA in enhancing inter-rater reliability (IRR) measures. Future research should aim to further refine the parameters of GA and explore additional optimization techniques to improve IRR measures.