Volume 16, No 2, 2019
Feature Selection For Crop Yield Prediction Using Optimization Techniques
Chellammal Surianarayanan , Kodimalar Palanivel
Abstract
Agriculture sector forms the basis of Indian economy. Prediction of yield of crops is an important issue in agriculture as the domain is facing a huge challenge of producing effective crops having high yield while maintaining the sustainability of natural resources. In addition, early prediction of yield enable the farmers to take precautionary actions to improve the productivity of crops. Crop yield depends on various factors including weather parameters, soil parameters, irrigation availability, fertilizer’s needs, farm capacity, etc. In general, early prediction is being done by analyzing the above data which is archived over several years using machine learning techniques. Identifying relevant and more useful attributes from the set of all available attributes(called feature selection) which really affect the yield becomes important before performing the prediction process using machine learning algorithms. The focus of this work lies with feature selection. In this work, four different feature selection techniques, namely, Binary Cuckoo Search(BCS), Relief, Grey Wolf Optimization(GWO) and Principal Component Analysis(PCA) have been employed over the agricultural data collected from field and preprocessed using Monte Carlo method. The selected features are given for yield prediction using multiple regression. The accuracy of prediction obtained using the above four methods are presented. Results are discussed.
Pages: 292-307
Keywords: crop yield prediction, feature selection, Binary Cuckoo Search, Grey Wolf Optimization, Principal Component Analysis, machine learning techniques, multiple regression