Crop recommendation in precision agriculture: a systematic literature review of methods, trends, and challenges
TL;DR
The review showed that ensemble learning-based approaches, particularly Random Forest and Extreme Gradient Boosting (XGBoost), are powerful for predictive performance on various agricultural datasets and shows the importance of scalable and explainable AI systems for real applications in agriculture.
Abstract
Crop recommendation methods have become an essential component of modern agriculture, helping farmers identify the most suitable crops based on soil, climatic, and environmental conditions. As key applications of precision agriculture, these methods increasingly employ artificial intelligence (AI), particularly machine learning (ML) and deep learning (DL), to improve crop prediction, recommendations, productivity, and farming decision-making. This study presents a systematic literature review (SLR) of ML and DL techniques applied to crop recommendation and related agricultural applications. From 183 identified studies, 129 articles published between 2020 and 2026 were carefully selected through a structured screening process for comprehensive analysis. Relevant studies were retrieved from major academic literature databases and publishing platforms, including ScienceDirect, Scopus, SpringerLink, MDPI, Nature, Frontiers, IEEE, Wiley, and Google Scholar. The review used the PRISMA protocol and showed that ensemble learning-based approaches, particularly Random Forest and Extreme Gradient Boosting (XGBoost), are powerful for predictive performance on various agricultural datasets. Traditional ML approaches such as Support Vector Machines, Decision Trees, and k-Nearest Neighbors are still commonly used. At the same time, Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks are used for remote sensing and time-dependent agricultural analysis. The most frequently used dataset source is Kaggle, and typical inputs include soil nutrients (NPK), soil pH, weather conditions, and satellite indices such as NDVI and EVI. Many studies have achieved high accuracy, but most are based on static datasets, which reduces their reliability in real-world scenarios. The main research gaps are limited real-time deployment, low integration of multiple data sources, low cross-regional validation, and low model interpretability. The review shows the importance of scalable and explainable AI systems for real applications in agriculture.