Application of Unsupervised Learning and Data Clustering to Saudi Online Stores

We applied data mining techniques to analyze customer behavior for an online Saudi store using historical purchase data. Unsupervised learning was used to segment customer behaviors using the following clustering algorithms: k-means clustering, hierarchical clustering,and DBSCAN. Clustering results were compared on the basis of the Silhouette score, Davies-Bouldin index, and Calinsk-Harabasz index. The results revealed that k-means clustering with $\mathrm{k}=5$ produced the most logical clusters. The clustering techniques managed to segregate customers into different segments according to total payments and the number of purchases they made. The clusters showed slight differences in customer spending habits. This approach had potential merit for the online store, as it makes it easy to target marketing for each category, such as presenting discount coupons, introducing new products and services, and making changes to existing services according to customer requirements.