Application of K-Means and Partitioning Around Mediods (PAM) clustering techniques on Maize and Beans yield in Tanzania
No Thumbnail Available
Date
2016
Journal Title
Journal ISSN
Volume Title
Publisher
KY Publications
Abstract
This paper contributes to the application of k-means and k-mediods multivariate statistical methods for the purpose of revealing optimal clusters and assessing the consistency of individual districts within the group. Data used were extracted from united republic of Tanzania (Ministry of Agriculture, Livestock and Fisheries(MALF) (2003/12) consisted a total of (n=36 districts) with both maize and beans yield. The R-statistical computing version (3.1.1) was used. The study findings revealed that 36 districts were grouped into six clusters using k-means algorithm. Using the k-mediods, it was revealed that only 11 districts were found to very well structured since their silhouette width (Si) is above 0.5. Nevertheless, the clusters validation was done in such a way that individual district whose silhouette width (Si) close to 1 was regarded as highly consistency clustered whereas districts with (Si) greater or equal to 0.25 were said to be somehow well clustered and otherwise. The paper concludes that few districts that are very consistent given the threshold margin should to be monitored and evaluated effectively to ascertain productivity. The study recommends that the government should pay attention on allocating the scarce resources to the consistency clusters along with policy review in favour of smallholder farmers through access and timely for all important farm inputs in future.
Description
Article published by KY Publications in the Bulletin Of Mathematics And Statistics Research Vol.4.Issue.4. p. 146-158
Keywords
Clustering, k-means and PAM or k-mediods, k-methods
Citation
APA