An Efficient Crop Yield Prediction Framework Using Hybrid Machine Learning Model

Abstract

Given India's vast expanse and dense population, the prediction of agricultural yields is crucial for ensuring food security. The task, however, is complex due to the influence of a multitude of factors, such as agricultural practices, environmental conditions, and technological advancements. Existing machine learning (ML) models face difficulties due to the quality and variability of data, model overfitting, intricate model structures, insufficient feature engineering, and temporal dependencies. Therefore, a robust and efficient model that addresses these challenges is imperative. In this study, an investigation was conducted using five prevalent ML algorithms -Random Forest (RF), XGBoost, Decision Tree (DT), Support Vector Machine (SVM), and Linear Regression (LR)on a crop prediction dataset sourced from Kaggle. Algorithms that exhibited the highest coefficient of determination (R² ) were selected to construct a hybrid model for aggregate prediction. Results demonstrated that the proposed hybrid model, encompassing DT, XGBoost, and RF, surpassed individual classifiers in terms of R² score and outperformed the existing models, achieving an accuracy of 98.6%. This provides a robust and efficient framework for crop yield predictions. Consequently, a user-friendly tool, 'Crop Yield Predictor', was developed, rendering the model accessible and practical for on-ground applications in agriculture. This tool effectively translates complex data and algorithms into actionable insights, bridging the gap between advanced machine learning techniques and practical agricultural applications.

A Systematic Review of Machine Learning Prediction Models for Colorectal Cancer Patient Survival Using Clinical Data and Gene Expression Profiles

“…The depth, architecture, and hyperparameters of neural networks, or the combination strategies in ensemble methods, can significantly influence the model's performance. Moreover, the way these techniques are applied, including the handling of imbalances in the dataset, feature selection, and optimization strategies, can also sway the results [43,44].…”

Artificial Bee Colony Algorithm-based Feature Selection and Hybrid ML Framework for Efficient Rice Yield Prediction

“…Filter, Wrapper, and embedded methods are some of the feature selection methods used by [5,9], helping farmers make more informed decisions about crop management and improving yields. Authors in [31] used Feature shuffling and Feature performance feature selection methodology with a hybrid model comprising DT, XGBoost, and RF achieving a coefficient of determination (R 2 ) of 98.6. Filter methods evaluate features independently of the classification model and rank them based on their correlation or mutual information with the target variable [6].…”

“…Wrapper methods evaluate feature subsets by repeatedly training and evaluating a classification model on different subsets. They search for the optimal subset of features that gives the best model performance but can be computationally expensive [31]. Embedded methods integrate feature selection into the model training process itself [32].…”