Comparative Analysis of Soil Classification Techniques to Predict Fertility Index for Kolhapur District, Maharashtra

Abstract

Agriculture is a major contributor to the Indian economy. Various factors like soil, weather, rain, fertilizers, pesticides, at all play a vital role in determining crop yields. Soil is one of the important factors for improving total crop productivity. We proposed an approach for predicting fertility by analysing characteristics of soil using data mining (DM) techniques. In the current study, we used soil data from the year 2019–20 with 7099 samples in the dataset. This data includes properties like macronutrients (nitrogen, phosphorus, potassium and sulphur), and micronutrients (zinc, iron, etc.). Using these properties, the fertility index (FI) was predicted using DM classification algorithms like J48, Naive Bayes, JRip, PART and REPTree. The results showed that the J48 algorithm gave better accuracy (99%) as compared to other classification algorithms. The farmers and decision makers can determine the soil fertility index using the decision tree form provided by the J48 algorithm, and also this study can be used to suggest different fertilizers based on the nutrients detected in the soil sample.