Evaluation of Different Clustering Techniques in Classifying the Vegetable Growing Panchayats of Ernakulam District, Kerala

Abstract

The goal of this study was to evaluate different clustering techniques in classifying the vegetable growing locations of Ernakulam (EKM) district of Kerala so that same nutrient recommendation can be prescribed for panchayats coming under the same cluster. Hierarchical clustering (HC) and K –means clustering were performed to group the panchayats based on soil fertility status and thereafter comparison of various clustering procedures was done using Davies – Bouldin (DB) index. Different dissimilarity measures- Euclidean, squared Euclidean, Chebychev distance and Mahalanobis D2 were determined and single linkage, complete linkage and average linkage methods were adopted under these measures. The results revealed that Mahalanobis D2 was the better clustering procedure with seven clusters (DB index: 0.120) followed by average linkage method under Euclidean distance (DB index: 0.306) with seven clusters. Manjapra and Keerampara panchayats remained as individual clusters. Keerampara had strongly acidic soils (pH -5.17) with high available Mg (158 mg kg-1) while Manjapra soils had low Mg availability (19 mg kg-1) and high S content (57 mg kg-1). Kakkad, Kalady and Vengoor came under cluster I which possessed approximately same EC (0.15-0.19 dS m-1), OC (2-2.4%) and Mg (71-73 mg kg-1) content. Chengamanadu and Vengola came under cluster III while Ayyampuzha and Mudakkuzha belong to cluster IV.