Wholesale Vegetable Price Profiles in Assam Using Nearest Neighbour and Average Linkage Clustering
Subhalakhmi Changmai *
Jagannath Barooah University, Jorhat, Assam, India.
*Author to whom correspondence should be addressed.
Abstract
Aims: To identify interpretable combinations of everyday wholesale vegetable prices in Assam and compare nearest-neighbour and average-linkage clustering.
Study Design: Exploratory observational study using publicly available secondary market quotations.
Place and Duration of Study: Sixteen reporting markets in Assam, India, from January to September 2026.
Methodology: The dataset contained 612 complete market–day observations of modal prices for potato, brinjal, tomato and cabbage. Prices were converted to rupees per kilogram and standardised. Single-linkage and average-linkage solutions with two to six groups were assessed. Supplementary checks used internal validation indices, 100 markets–week block subsamples, an unusual-quotation sensitivity refit and 135 weekly profiles.
Results: Single linkage showed pronounced chaining. Four-group average linkage produced groups of 394, 47, 169 and two observations, with a mean silhouette width of 0.428. The 169-observation group combined lower potato prices with higher brinjal, tomato and cabbage prices. Two groups achieved a higher silhouette of 0.599 but chiefly isolated two unusual potato quotations. Block subsampling gave a median adjusted Rand index of 0.892, with an interquartile range of 0.407–0.926. Weekly four-group clustering had a silhouette width of 0.448, although a singleton remained.
Conclusion: Average linkage provides an interpretable exploratory summary of price combinations. Uneven reporting and variations in stability prevent claims of a unique natural partition, causal seasonality or household affordability.
Keywords: Assam, wholesale vegetable prices, AGMARKNET, hierarchical clustering, nearest neighbour, average linkage, cluster validation