Proficient Possessions Supervision Using Various Clustering Techniques

Feed is the largest input in any livestock enterprise and the rapid increase in feed prices and shortage of feed resources has been one of the major constraints for farmers, livestock industries, planners and the policy makers. This calls for prudent management of available resources and application of computing techniques can be one of the possible potential approaches. India is endowed with a wide range of feed resources varying widely in their composition and utility for different livestock species. Clustering of feed resource into different groups based on the composition can help in better feed management. To evaluate and to suggest a best technique for clustering feed resources, we have evaluated three clustering techniques viz. K-means, spectral k-means and auto spectral on two different data sets containing 236 and 106 feed resources with major constituents like crude protein, crude fiber ash, fat etc., . The composition of commonly used feed resources (236 and 106) were sourced from the published articles in various scientific journals and based on the composition of the feeds, they were subjected to clustering techniques. A total of three clustering techniques – k means, spectral k means and auto spectral were employed and the outputs were compared against the grouping done by subject matter experts. The outputs of different clustering techniques were validated with the expert grouping using precision, recall and F-measure. Of the three methods used it was found that kmeans was the best and closest to the experts classification. Clustering of feeds into different groups based on their composition will form a sound basis to select different feeds for formulating economic and nutritionally balanced diets leading to better feed management.