Proportional Cox Regressive Maxout Deep Recurrent Percepted Neural Learning Classifier for IOT aware Soil Quality Prediction

Abstract

Analyzing soil quality is the most important task in the agriculture field since it origin for
further improving the crop yield. Considering the soil conditions as input parameters is to
predict the nutrient content in the soil. But the soil quality prediction is a great challenge
because land-atmospheric interactions are complex and diverse in space and time. In order to
perform accurate prediction, a novel technique called Proportional Cox Regressive Maxout
Deep Recurrent Percepted Neural Learning Classifier (PCRMDRPNLC) is introduced. The
main aim of the proposed PCRMDRPNLC technique is to improve the prediction accuracy
and minimize time consumption. First, the IoT is used for collecting data. After that, the
proposed PCRMDRPNLC technique performs two major processes namely feature selection
and classification. At first, the feature selection process is carried out to select the significant
features from the dataset using KMO proportional correlative cox regression-based feature
selection. This helps to reduce the time complexity of soil quality prediction. Followed by,
the classification is performed using Herfindahl Maxout Deep Recurrent multilayer
perceptive neural learning classifier that consists of cascaded subnetworks for deeply
analyzing the extracted features in the layers. Deep feature learning performs the final
prediction results by analyzing the testing and training features using Herfindahl–Hirschman
Index-based classification. Based on the classification results, the different soil qualities are
correctly predicted with minimum error. Experimental assessment is performed with different
quantitative metrics such as prediction accuracy, precision, recall, F-measure and prediction
time. The analyzed results demonstrate the superior performance of our proposed
PCRMDRPNLC technique when compared with the two state-of-the-art methods.