Balancing Soil Parameters and Farmers Budget by Feature Selection and Ordered Weighted Averaging

Abstract

A method is presented allowing farmers at Shiraz in the Fars province of Iran to balance in between their budget and the soil parameters. First, the three alternatives (Best-First, Greedy-Stepwise and Ranker) of the Feature Selection Method identify the most critical soil fertility parameters. Training data model evaluation indicate that the Greedy-Stepwise feature selection algorithm (with attribute evaluator of CFS-Subset-Eval) presents the highest accuracy for the particular study area. Soil fertility is found to highly depends on Potassium, Phosphor, and Organic Carbon while Copper, Iron, Manganese, and Zinc dependencies are rejected. Finally, by utilizing Ordered Weighted Averaging, six maps with different risk levels in terms of the soil fertility are constructed allowing alternative management options according to the farmers budget. The major scientific contributions are summarized to a) the identification of soil fertility parameters, and the b) construction of maps modeling soil fertility for various degrees of uncertainty allowing agricultural cost effective planning in the study area.