A hybrid machine learning model for optimized mixed-crop recommendation

Abstract

Farmers today encounter more challenges when selecting appropriate variety of crops depending on their farm soil nutrients and climate. This research will assist farmers in choosing suitable mixed-crops depending on the individual farms soil and climate conditions in Andhra Pradesh, India. Using the dataset sourced from Indian Institute of Soil Science (IISS), Bhopal with 2,552 entries. Previous studies focused on only single-crop recommendations. This work proposes a novel hybrid mixed-crop recommendation system (CRS) that incorporates several machine learning (ML) techniques comprise of random forest-ExtraTrees (RF-ExtraTrees), decision tree-C4.5 (DT-C4.5), extreme gradient boosting-gradient boosting (XGBoost-GBoost), quadratic discriminant analysis-linear discriminant analysis (QDA-LDA), and support vector machine-stochastic gradient descent (SVM-SGD) were utilized to recommend mixed-crops. To enhance the reliability of the training process, 20% of the dataset was held in reserve for validation to analyze model performance. The result of the proposed work shows that all the hybrid ML models applied were viable, and RF ExtraTrees has achieved 95.91% best accuracy, 95.08% precision, and 95.91% recall, when contrasted to the other ML models.