CART Algorithm in Soil Classification

Authors : K Aditya Shastry, Supriya Hebsur, Sanjay H. A

Abstract

As Agriculture is one of the basic and most done occupations in India, the farmers are the back bone for the development of the country. The farmers are heavily dependent on different weather conditions, fertilizer quality, soil types, etc In this regard, determining whether the soil type is problematic or non-problematic is important for crop growth. Indian Government is currently providing useful information to farmers based on manual methods which are error prone and also time consuming. As a first initiative towards this we have used a well known algorithm called CART algorithm for soil classification. In this work we have implemented CART algorithm which is based on decision tree for classifying the Soil Set into problematic and non-problematic soils by considering appropriate parameters of soil. We have obtained the soil data set from National Bureau of Soil Sciences located in Hebbal, Bangalore. We have compared the CART algorithm with the brute force method for soil classification methods used in the soil laboratories. With the available set of training data of 200 records we were able to achieve 100% accuracy.

Keywords : Data Mining, CART, Decision Tree, Agriculture, Soil