Machine Learning Ensemble Classifiers for Feature Selection in Rice Cultivars

Abstract

Machine Learning (ML) has a big impact on smart farming, especially rice productivity. This is especially true for intelligent farming. Machine Learning is crucial for seed prediction, germination, crop production, soil moisture, and land suitability evaluation. Selecting a rice cultivar requires considering local environmental and seed factors. This research examines classification algorithms like K-Nearest Neighbor (KNN), Decision Tree (DT), NaiveBayes (NB), Support Vector Machine (SVM), and Random Forest (RF) with wrapper feature selection techniques like SFFS, SBEFS, CBFS, VIF, and RANDIM for environmental and seed data. This research study proposes the best feature selection approach with classifiers for the environmental & seed factor dataset to select the rice cultivar. This study collected 10 rice cultivar characteristics (short, medium, and long duration, average yield, minimum and maximum days, seed types, grain weight, grain type, color, parentage, and other special characteristics) and 7 environmental characteristics (Starting Month, Ending Month, Rainfall Actual, Rainfall Normal, Temperature Minimum, Temperature Maximum, District) from 2002 to 2022 in Madurai, Tamilnadu. The variance inflation factor (VIF) of the wrapper feature selection approach with decision tree classification algorithm yields 99.63% accuracy and 4.3% error rate compared to other classification algorithms and wrapper feature selection techniques.