Performance Collation of Machine Learning Algorithms Intent to Classify Agriculture Data of Rajasthan State, India

Abstract

Soft computing technique is used to develop a computational intelligent tool for precision agriculture. This paper mainly aimed to present a highly accurate model, to precisely predict the best crop founded under specific environmental conditions. Support vector machine (SVM), ZeroR, Naivebayes, BayesNet, RandomForest, lBk and decision tree (J48) are the well-liked data mining techniques that are used in this work for agricultural modelling and prediction. By applying machine learning to sensor information, cultivate administration frameworks are advancing into ongoing artificial insight empowered projects that give rich proposals also, bits of knowledge for agriculturist choice help and activity.
This paper presents the results of model build to predict the crop over nine subsequent years agricultural data of Rajasthan state, India. Data of 17 attributes and 2304 instances are pre-processed for better performance. Accuracy, kappa statistics, time taken and Area under Curve are estimated. The results obtained from considered models are compared and concluded that RandomForest outperformed at all the considered aspects. Feature engineering is crafted for advancing the results.