Machine learning model for automation of soil texture classification

Abstract

Soil formation is a long term process and diverse soils are formed in different localities due to various soil forming factors over the landscape. Soil classification plays critical role in various aspects of agricultural engineering. Physico-chemical parameters play an important role in soil classification. In this paper, we present a comprehensive classification model for soil texture classification by using Linear Discriminant Analysis (LDA). We took the Physico-chemical properties of the soil, which include soil moisture, temperature, electrical conductivity, pH, organic carbon, available nitrogen, available phosphorus and potassium as independent variables, while the soil type was taken as the dependent variable. Feature selection is employed using Boruta algorithm. The performance of the proposed classification model is evaluated and expressed in terms of overall accuracy and kappa coefficient. Results show that the average prediction accuracy and kappa coefficient of the proposed model are 96.3% and 0.944 respectively, indicating that the model can be used effectively for soil classification for a set of suitable dependent variables.