Sensor based Framework for Soil Nutrients Prediction using Deep Learning Techniques

Abstract

A thriving economy of the country depends on agriculture. Soil is necessary for producing food in a sustainable manner as farming gets more intensive and demand rises, the quality of the soil may decrease. Smart soil prediction provides precise information on soil nutrient distribution needed for precision agriculture. Deep learning and machine learning techniques are currently the main drivers of intelligent soil prediction systems. This paper presents a hybrid deep learning based framework for identifying the soil nutrients in the sensor collected data and to predict the quality of the soil. The soil data collected from the sensors will be fed into the binary classifiers to classify the various soil quality then it will be fed into multi-class classifier to classify the soil nutrients. This is implemented to show the better suitability for any agriculture field to check its soil quality and lack of nutrients in it. The proposed one dimensional convolutional neural network-LSTM framework is compared with benchmark techniques like Support Vector Machine, K-Nearest Neighbors, Random Forest, Long Short Term Memory, One-Dimension Convolutional Neural Network, and Artificial Neural Network schemes. Our proposed model shows 98% accuracy in the prediction of soil quality compared to the traditional ones.