Multi-Variate Attribute Selection for Agricultural Data

Abstract

Farmers always have been concerned about the quantity of crops (yield) as well as the quality of crops (sugar content of the sugar beets). The quality and quantity of crops are affected by various attributes, some are natural elements (rain, sunshine etc) and some are not (the amount of fertilizer, seed type etc). Some techniques have been developed to discover attributes that are important to different crops’ yield. But within those selected attributes, how can we tell one attribute is more important than the other? The proposed algorithm is aimed to utilize the advantages of multiple response attributes to select the important attributes and then put the selected attributes in a hierarchical order. Although at the end this paper only focuses on yield prediction, any other target attribute can be a candidate for the prediction model.