Table 2 Summary of related work on machine learning and IoT applications in precision agriculture.

¹⁵

Multimodal ML for crop recommendation and yield prediction

Used equilibrium optimizer (EO), kernel extreme learning machine (KELM), and random forest (RF). Simulations and benchmark evaluations were conducted

Achieved 97.91% accuracy in reliable wireless environments. Highlights ML potential in improving crop management and yield

High accuracy and suitability for reliable wireless environments

Requires testing in resource-limited or real-world agricultural scenarios

²¹

Crop prediction using IoT and ML

Analyzed data from IoT architectures with 2200 cases and 8 attributes. Utilized decision table classifiers and multilayer perceptron for classification

Achieved 98.22% accuracy. Demonstrated robust model for precision farming using IoT and ML for crop prediction

Effective integration of IoT for crop prediction and fertilizer recommendations

Limited scalability to larger datasets and field conditions

¹⁰

ML-based cloud platform for crop recommendation

Compared KNN, DT, RF, XGBoost, and SVM for building a cloud-based recommendation engine

Proposed a cost-free, open-source platform for precision farming technologies to enhance acceptance in agriculture

Provides a scalable and accessible cloud-based solution for farmers

Limited discussion on real-time implementation and adoption barriers

¹⁸

Deep learning for crop and water quality assessment

Proposed a deep learning model considering solar exposure, humidity, soil pH, and water quality. Compared with SVM

Achieved 97% and 96% accuracy for crop and water quality predictions. Demonstrated ML potential in advancing agricultural productivity

High accuracy for water and crop quality assessment with multi-feature inputs

Lacks exploration of real-time or dynamic environmental changes

²²

Time-series crop yield modeling

Used datasets across fields and years with auxiliary data (e.g., soil conductivity, rainfall, MODIS images). Random Forest models developed for different growth stages

Predicted crop yields using spatially distributed data. Highlighted benefits of integrating time-series data for improving predictive models

Demonstrates the use of spatial and temporal data integration for yield prediction

Limited application to specific crop types and regions

²³

IoT-based NPK sensor for soil nutrient monitoring

Designed NPK sensor with fuzzy logic for nutrient deficiency detection. Integrated data into Google Cloud database with a Raspberry Pi 3 prototype

Demonstrated effective nutrient monitoring with alerts for farmers. Highlighted IoT-based solutions for improving soil management practices

Affordable and portable solution for real-time nutrient monitoring

Needs scalability for larger fields and better integration with other agronomic data

²⁴

Hybrid ML for yield prediction

Proposed a two-tier ML model with adaptive k-Nearest Centroid Neighbor Classifier (aKNCN) and Extreme Learning Machine (ELM). Used metrics like RMSE, R², and MAE

Improved accuracy and error reduction using hybrid ML models. Showcased IoT’s role in data-driven decisions for agricultural profitability

Combines feature selection and ML techniques for improved prediction accuracy

Lack of validation across diverse geographic or climatic conditions

²⁵

Nutrient profiling of mulberry leaf plantations

Employed Extreme Learning Method (ELM) with alternative activation functions. Classified soil traits like potassium, phosphorus, and boron across zones

Provided key insights into soil fertility and nutrient dynamics in Tamil Nadu, contributing to sustainable agricultural management

Highlights regional nutrient dynamics for sustainable agriculture

Limited scalability to non-mulberry crops and other regions

²⁶

Smart irrigation and IoT in agriculture

Developed a smart irrigation system using Agriculture 4.0 concepts, incorporating soil and climate data for decision-making

Presented a cost-effective urban agriculture solution for localized climate monitoring and decision support

Cost-effective system for urban agricultural monitoring

Needs validation in large-scale and rural agricultural environments

²⁷

CNN-RNN framework for crop yield prediction

Combined CNN and RNN for multimodal deep learning. Applied to decades of maize and soybean yields in the USA

Achieved better performance than Random Forest and other models. Demonstrated deep learning’s capability in large-scale yield prediction

Shows the strength of combining CNN and RNN for yield prediction in large datasets

Limited applicability for smaller farms and non-cereal crops