The implementation of K-Means clustering in kovats retention index on gas chromatography

Abstract

In this study, the retention index data of 146 compounds that are found in coal and petroleum-derived liquid fuels were grouped using the K-means clustering method, and the similarities between each cluster were analyzed. The psycho-chemical properties of each compound in the cluster were identified and compared with other clusters. Each compound’s retention index is grouped based on the similarity between the column polarity and heating rate of one compound to another. Based on the results of tests carried out on nine differentk values, it is known that the grouping with the value of k = 3 is the best determined from the obtained silhouette score = 0.568, where this score is higher than the score obtained on the other k values. The results of clustering with k = 3 obtained three clusters, namely cluster C1, cluster C2, and cluster C3. Cluster C1 and cluster C2 consist of chemical compounds that have a relatively low carbon number and molecular mass, but in cluster C2 the molecular mass of the compound is lower than in cluster C1. In contrast, the C3 cluster consists of chemical compounds that have a relatively high carbon number and molecular mass.