A dynamic multi-class Neupper classification for multiple crop yield prediction

Abstract

In agricultural applications, the most essential task is predicting the crop yield to classify the yield productivity over a certain interval of harvesting. The state-of-the-art classifiers are used for predicting the yield quality of any one crop whereas it takes more time to simultaneously train multiple types of crops for predicting their yield quality. For a specific crop yield prediction using soil parameters, a Krill-Herd (KH)-based feature selection with Dynamic Neupper (DNeupper) rule-based classifier has been proposed. However, multiple types of crops were not simultaneously predicted within a single classifier since it creates a multi-class classification problem. Hence in this article, KH with Dynamic Multi-Class Neupper (KHDMCNeupper) rule-based classification algorithm is proposed to predict all three crops such as rice, wheat and maize together with increased prediction accuracy. In this model, the most optimal soil parameters for all crops and their relative weights are computed based on KH and Rough Set (RS) theory. Then, these weight values are combined with soil parameters and given as input to the Artificial Neural Network (ANN) which is used to construct a tree in DNeupper classifier. By constructing a tree, the classification rules for all three crops are generated to predict the yield quality. Thus, the proposed classification technique can support simultaneous prediction of multiple crops with high accuracy. Finally, the experimental results show the efficiency of the KHDMCNeupper classifier compared to the KHDNeupper classifier in terms of accuracy, precision, recall and f-measure.