Grouping of Medicinal Drugs Used for Similar Symptoms by Mining Clusters from Drug Benefits Reviews

Abstract

Clustering is an unsupervised process for partitioning a set of data into a set of meaningful sub-classes or sub-datasets, called clusters. In this paper, the k-means clustering algorithm is used on a textual dataset of unlabeled reviews of medicinal drugs in order to group the drugs with similar usage and benefits. Two alternative clustering algorithms, agglomerative clustering and BIRCH (balanced iterative reducing and clustering using hierarchies), are used on the same dataset for comparison. The methodology for text preprocessing and feature extraction is explained. The results are analysed using quantitative performance metrics (Silhouette Score and Calinski-Harabasz Score), cluster sizes, as well as visualization by the means of scatter plot of clusters, distances from final centroids, word clouds and dendrogram.

Keywords: Drug Benefit Review, k-means Clustering, Hierarchical Clustering, Agglomerative Clustering, BIRCH, Silhouette Score, Calinski-Harabasz Score