APPLICATION OF K-MEANS AND K-MEDOID CLUSTERING TECHNIQUES ON RICE YIELDS IN NIGERIA

Abstract

This paper explores the application of two clustering algorithms, K-means and K-medoid to classify rice yields across 36 Nigerian states and the Federal Capital Territory (FCT). The study aims to identify natural groupings within the data to classify the states base on their similarity in the rice yields. The findings reveal that the K-means and K-medoid clustering effectively grouped rice and maize yields into six clusters each. The silhouette width indicated that K-medoid out performed K-means in cluster quality with an average silhouette width of 0.42 and for rice compared to K-means with silhouette width of 0.39 and. The analysis highlighted significant clusters, such as clusters three and five for rice using K-means and K-medoid which represent regions with similar average crop yields. These insights underline the importance of targeted governmental interventions to improve agricultural productivity by focusing on areas with average yields. The results also suggest that strategic investments in infrastructure, agricultural inputs, and policy reforms are crucial for boosting productivity and reducing poverty. Overall, the paper concludes that K-medoid is the superior clustering technique, delivering the highest silhouette width and the most accurate classifications for both crops, with the fewest misclassifications. This research provides a valuable framework for regional agricultural planning and resource allocation in Nigeria. The paper recommends that the government should pay attention on allocating the scarce resources to the consistency clusters along with policy review in favor of smallholder farmers through access and timely for all important farm inputs in future.