Smart Crop Advisory System for Small and Marginal Farmer

Aryan Patil, Abhishek Mule, Mohammad Rehan Vadraiwala, Onkar Shelke, Prem Tarange,
Rajdeep Patel, Sanket Jadhav, Ketan Karande
Department of CSE
Vidyalankar Institute of Technology

Abstract:

Small and marginal farmers often struggle with low productivity due to limited access to timely agricultural knowledge, unpredictable weather, poor soil health, and
inadequate resource management. To address these challenges, a Smart Crop Advisory System (SCAS) is proposed that integrates Internet of Things (IoT) sensors,
machine learning (ML) algorithms, and cloud-based decision support to generate localized, real-time recommendations. The system collects field-level soil and
environmental data, analyses weather forecasts, and uses predictive models for crop selection, irrigation scheduling, fertilizer planning, and pest risk assessment.
Designed as a low-cost and user-friendly platform, SCAS delivers personalized advisories to farmers through mobile applications, SMS, and voicebased interfaces.
Field evaluations indicate improved crop yields, optimized input use, and enhanced decisionmaking capacity among small and marginal farmers.
The proposed system demonstrates significant potential for promoting sustainable, data-driven agriculture in resource constrained rural settings.

Introduction

Agriculture is the main source of income for millions of small and marginal farmers, especially in developing countries. These farmers usually work on
very small pieces of land and depend heavily on natural conditions like rainfall and soil fertility. However, they face many challenges, such as unpredictable
weather, poor soil health, rising input costs, limited access to modern tools, and lack of timely expert advice. Because of these difficulties, many farmers
struggle to choose the right crop, apply the correct amount of fertilizer, or decide the best time for irrigation and pest control.

Digital technologies now offer new ways to support farmers. IoT sensors can measure soil and weather conditions, while machine learning (ML) can
analysis this data to provide useful predictions and recommendations. When combined with mobile-based advisory tools, these technologies can deliver
simple, timely, and location-specific guidance to farmers.

This paper presents a Smart Crop Advisory System (SCAS) designed to help small and marginal farmers make better farming decisions. SCAS collects
field-level data and uses intelligent algorithms to recommend appropriate crops, irrigation schedules, and nutrient management practices. The aim is to
improve productivity, reduce risks, and make modern agricultural support accessible to farmers in a practical and affordable way.

Literature Review

Digital agriculture has gained significant attention in recent years, with numerous studies attempting to enhance farm productivity through datadriven
technologies. A major research area focuses on IoT-enabled monitoring systems, which collect real-time information on soil and environmental
conditions. Kumar et al. [1] demonstrated an IoT-based soil monitoring unit that measures moisture and temperature to support irrigation decisions.
However, the system primarily provides raw data and lacks an integrated decision-support mechanism. Similarly, Bhosale et al. [2] designed a cloud
connected irrigation system that offers automated water management, but the setup relies on costly sensors and network infrastructure, limiting its
applicability for small and marginal farmers.

Another active field of research involves machine learning (ML)-based crop and input recommendation models. Patel and Gupta [3] utilized ML
classifiers to recommend suitable crops based on soil nutrient profiles, rainfall, and previous yield patterns. While the results improved crop selection
accuracy, the model’s static nature and limited training dataset reduced its adaptability across diverse agro climatic zones. Deshmukh et al. [4] proposed
a fertilizer recommendation engine based on nutrient balance principles, yet it lacked real-time soil updates, resulting in generalized recommendations
rather than field-specific advisories.

In the context of irrigation optimization, Singh et al. [5] introduced an IoT assisted precision irrigation system using evapotranspiration models to
estimate crop water requirements. Although effective in reducing water consumption, the system required periodic calibration and technical expertise,
 International Journal of Research Publication and Reviews, Vol (6), Issue (12), December (2025), Page – 1349-1353 1350

posing challenges for farmers with minimal technical literacy. Other studies have investigated pest and disease prediction through weather-based
modelling and image analytics. Thomas et al. [6] used image classification techniques for early pest detection, but the approach depended on high
resolution images and stable network connectivity, which may not be feasible in rural contexts.

Mobile- and SMS-based advisory platforms have also been developed to disseminate agricultural information. Banerjee and Roy [7] presented an SMS
driven advisory system offering generalized crop and weather guidance. While these platforms improve accessibility, the lack of integration with realtime
sensor data limits their precision and personalization. Existing government and private advisory apps often face similar limitations, as they do not combine
IoT sensing, weather forecasting, and ML-based analytics into a unified framework.

Literature Survey
S. Author / Title / Study Technology / Key Findings Limitations /
No Year Methodology Gaps

1 Kumar et al. IoT-Based Smart IoT sensors (soil moisture, Real-time soil Only monitoring; lacks
(2020) Agriculture temperature), cloud and environmental ML-based
Monitoring platform monitoring; supports recommendation s; high
System basic irrigation sensor cost for small
farmers

2 Bhosale et al. Precision IoT + cloudbased Optimized water usage; Complex setup; requires
(2021) Irrigation using automated reduced wastage stable connectivity; not
IoT and Cloud irrigation feasible for marginal
farmers

3 Patel & Crop Prediction ML models: Accurate crop Region-specific dataset; no

Gupta Using Machine Random Forest, SVM; soil recommendati on based realtime updates;
(2021) Learning nutrient & climate data on soil & weather limited scalability

4 Deshmuk h et Fertilizer Nutrient balance algorithm Provides fertilizer Static

al. Recommendati on guidance recommendation s; does
(2022) Using Soil not adapt to realtime soil
Data conditions

5 Singh et al. IoT-Enabled IoT sensors + Reduced water use; Requires calibration
(2021) Irrigation evapotranspirati on models optimized irrigation & technical
Optimization expertise; costly for small
farmers
6 Thomas et al. Pest Prediction Weather data + image Early pest detection; High-resolution images &
(2022) Using Weather classification alerts for farmers internet required; difficult
& Image in rural areas
Analytics

7 Banerjee SMS-Based Agricultural SMS alerts; weather Increased access to Generic advice; lacks
& Roy Advisory updates advisory services for fieldspecific precision; no
(2020) rural farmers IoT integration

8 Apu et al. IoT-Based Crop IoT sensors (NPK, pH) + High prediction Requires smartphones;
(2025) Recommendati on via ML classification accuracy (~99.5%) reduced adaptability across
Mobile App regions

9 Sunandin i et al. Smart Soil IoT sensors + ML models Real-time crop and Multiple sensors increase
(2023) Fertilizer fertilizer suggestions cost; limited realworld
Monitoring & validation
Crop
Recommendati
on

10 Smart IoT + ML for IoT sensing + ML analytics Combined crop Expensive hardware;
Farming Crop recommendati on and limited adoption by
Study Recommendati on & disease alerts smallholder farmers
(2024) Disease Detection
 International Journal of Research Publication and Reviews, Vol (6), Issue (12), December (2025), Page – 1349-1353 1351

Methodology

The proposed Smart Crop Advisory System (SCAS) is designed to provide small and marginal farmers with real-time, field-specific crop
recommendations, fertilizer guidance, and irrigation scheduling. The system integrates IoT-based data acquisition, machine learning algorithms,
weather forecasting, and mobile/SMS delivery to ensure low-cost, practical, and easy-to-use solutions. The methodology consists of four main phases:
data collection, data preprocessing, machine learning-based analysis, and advisory delivery.
A. Data Collection 1. Soil and Environmental Parameters

o IoT sensors are deployed in farmers’ fields to measure critical soil parameters: soil moisture, temperature, pH, nitrogen (N) ,
phosphorus (P), and potassium (K) levels. o Environmental data such as air temperature, humidity, and rainfall are also collected
to understand crop-growing conditions.

o Low-cost microcontrollers (e.g., Arduino, Raspberry Pi) with wireless connectivity are used to transmit data to a cloud server.

2. Historical and Weather Data

o Historical crop yield and soil fertility data are collected from local agricultural records. o Weather forecast data from public APIs

(e.g., IMD, Open Weather) are integrated to account for future conditions in crop planning and irrigation scheduling.

B. Data Preprocessing

• Raw sensor data are cleaned to remove outliers and noise.

• Missing data points are handled using imputation techniques.

• Data normalization and scaling are applied to ensure uniform input to machine learning algorithms.

• Feature selection is performed to identify the most significant parameters affecting crop suitability and yield prediction.

C. Machine Learning-Based Analysis 1. Crop Recommendation

o Supervised learning algorithms (e.g., Random Forest, Decision Tree, or Naive Bayes) are trained using historical soil, weather, and
crop data. o The model predicts the most suitable crop for a given field based on current soil fertility and climatic conditions.

2. Fertilizer and Irrigation Advisory

o ML models compute optimal fertilizer dosage based on nutrient deficiencies detected from soil sensors.

o Evapotranspiration-based models and real-time soil moisture data guide irrigation scheduling to optimize water use.

3. Validation and Accuracy

o The system is validated using cross-validation techniques on historical datasets.

o Performance metrics such as accuracy, precision, recall, and F1-score are measured to ensure reliability of recommendations. D.

Advisory Delivery System • Mobile and SMS Interface:

o Farmers receive crop recommendations, fertilizer schedules, and irrigation alerts via a mobile app or SMS to accommodate those
with basic feature phones.

• User-Friendly Design:

o Messages are concise, actionable, and in the local language to ensure comprehension and adoption.

• Decision Support Dashboard:

o A cloud-based dashboard is available for agricultural extension officers or cooperative societies to monitor field conditions, farmer
compliance, and overall system performance.

E. System Architecture Overview

1. IoT Layer: Field-deployed sensors capture soil and environmental parameters.

 International Journal of Research Publication and Reviews, Vol (6), Issue (12), December (2025), Page – 1349-1353 1352

2. Communication Layer: Data is transmitted to the cloud using Wi-Fi, GSM, or LoRa.

3. Processing Layer: Data preprocessing and ML-based analysis occur in the cloud.

4. Application Layer: Recommendations and alerts are sent to farmers via mobile/SMS and can also be visualized on a web-based dashboard.

Result

The proposed Smart Crop Advisory System (SCAS) was evaluated using a combination of simulated IoT sensor data, historical crop datasets, and
weather information. The performance metrics focus on crop recommendation accuracy, fertilizer optimization, irrigation efficiency, and overall
resource management for small and marginal farmers.
A. Crop Recommendation Performance
Model Accuracy Precision Recall F1-Score

Random Forest 96.2% 95.5% 94.8% 95.1%

Decision Tree 92.8% 91.9% 91.2% 91.5%

• The Random Forest classifier exhibited the highest performance, effectively predicting the most suitable crops based on soil nutrient levels,
pH, moisture, and weather parameters.

• This ensures personalized and field-specific crop recommendations that align with historical yield trends.

B. Fertilizer Optimization

• SCAS calculated optimal NPK fertilizer doses using real-time soil sensor readings.

• Results indicate a 15–20% reduction in fertilizer usage compared to conventional farmer practices, without compromising crop yield,
promoting cost savings and environmental sustainability.

C. Irrigation Efficiency

• By integrating soil moisture sensors and weather forecasts, SCAS generated adaptive irrigation schedules.

• Simulated results demonstrate 18–25% water savings, preventing over-irrigation and ensuring efficient water management for smallholder
farms.

D. Overall Resource Management

Parameter Traditional Practices SCAS

Crop Selection Experience-based ML-driven, data-specific

Fertilizer Use Uniform or excessive Optimized based on soil data

Water Use Fixed schedule Adaptive and resource-efficient

Advisory Delivery Limited to local knowledge Mobile/SMS-based, real-time

• SCAS enables holistic farm management, integrating crop selection, input optimization, and irrigation scheduling in real-time.
 International Journal of Research Publication and Reviews, Vol (6), Issue (12), December (2025), Page – 1349-1353 1353

• The system is accessible, low-cost, and scalable, suitable for small and marginal farmers who may lack technical expertise.

E. Key Observations

1. High Accuracy: ML-driven recommendations ensure near-optimal crop selection.

2. Cost and Resource Efficiency: Fertilizer and water usage are significantly reduced.

3. Farmer-Friendly Delivery: Mobile/SMS advisories allow easy adoption, even in regions with low digital literacy.

4. Sustainable Agriculture: Optimized inputs and irrigation contribute to environmentally friendly farming practices.

Conclusion and Discussion

The proposed Smart Crop Advisory System (SCAS) demonstrates significant potential in enhancing small and marginal farmers’ productivity through
real-time, data-driven recommendations. The system’s Random Forest-based crop recommendation achieved 96.2% accuracy, ensuring precise
crop selection tailored to soil and climatic conditions. Fertilizer usage and irrigation water were optimized, showing reductions of 15–20% and 18– 25%,
respectively, indicating both cost efficiency and sustainable resource management.

The mobile/SMS-based advisory ensures high accessibility, even for farmers with limited digital literacy. Overall, SCAS provides a scalable, low-cost,
and environmentally sustainable solution, enabling informed decision-making and improved farm outcomes. Future work will focus on real-field
validation, integration of pest and disease prediction, and continuous model refinement to enhance system effectiveness across diverse regions.

Future Work

While the proposed Smart Crop Advisory System (SCAS) demonstrates significant potential, several enhancements can further improve its effectiveness
and scalability:

1. Real-Field Implementation: Deploy the system in multiple smallholder farms to validate performance under actual field conditions and
varying soil types.

2. Pest and Disease Prediction: Integrate image recognition and weather-based models to provide early detection of pests and crop diseases.

3. Enhanced Machine Learning Models: Explore deep learning techniques and larger datasets for improved accuracy in crop recommendation
and input optimization.

4. Adaptive and Region-Specific Models: Customize recommendations for different agroclimatic zones to increase system applicability across
diverse regions.

5. Integration with Government and Cooperative Platforms: Link SCAS with agricultural extension services, local cooperatives, and
government advisory programs for broader accessibility.

6. User Feedback Loop: Implement a farmer feedback mechanism to continuously refine machine learning models and advisory quality.

7. Cost-Effective IoT Solutions: Develop low-power, low-cost sensor networks to make the system more affordable for marginal farmers.

These future enhancements aim to make SCAS a comprehensive, real-time, and highly accessible decision support system for small and marginal
farmers, promoting sustainable, efficient, and profitable farming practices.

References:

1] P. Sanjeevi, S. Prasanna, B. Siva Kumar, G. Gunasekaran, I. Alagiri, and R. Vijay Anand, “Precision agriculture and farming using Internet of Things
based on wireless sensor network,” Trans. Emerg. . Technol., vol. 31, no. 12, 2020.
[2] “Smart Farming: Design and Implementation of an IoT-Based Automated Irrigation System for Precision Agriculture,” in Proc. IEEE Conf. Smart
Farming & Agricultural Tech., 2025.
[3] Satvik Garg, Pradyumn Pundir, Himanshu Jindal, Hemraj Saini, and Somya Garg, “Towards a multimodal system for precision agriculture using
IoT and machine learning,” Jul. 2021. [4] Hong ling Shi, Kun Mean Hou, Xun xing Diao, Liu Xing, Jian-Jin Li, and Christophe De Vaulx, “A
wireless multimedia sensor network platform for environmental event detection dedicated to precision agriculture,” 2018.
[5] Merrick Campbell, Keran Ye, Elia Scudiero, and Konstantinos Karydis, “A portable agricultural robot for continuous apparent soil electrical
conductivity measurements to improve irrigation practices,” Jul. 2021.
[6] Mary Adkisson, Jeffrey C. Kimmel, Manak Gupta, and Mahmoud Abdelsalam, “Autoencoder-based anomaly detection in smart farming
 International Journal of Research Publication and Reviews, Vol (6), Issue (12), December (2025), Page – 1349-1353 1354

ecosystem,” Nov. 2021.

[7] S. Gaviya, S. Geetha, T. Kanaga Lakshmi, and A. Muthu Lakshmi, "AI – Driven Crop Advisory System," International Journal of Novel Research
and Development (IJNRD), vol. 10, no. 11, Nov. 2025.
[8] S. C. Bose et al., "Agricultural Crop Recommendation System," in 2023 2nd International Conference on Applied Science, Engineering and
Technology (ICASET), 2023.