Crop Prediction Using Sensors and Machine Learning

Abstract

The agricultural industry is rapidly embracing emerging technologies to enhance crop forecasting and resource utilization. This project suggests an intelligent crop forecasting system based on IoT and machine learning for sustainable agriculture. Crop selection is usually done based on guesswork by the farmers, resulting in low harvest and wasted resources. Crop yield forecasting before harvesting is also a severe problem in developing countries. The system relies on Arduino-based hardware interfaced with sensors for measuring temperature, humidity, rain, soil moisture, and water levels. The values are applied to train a logistic regression model for crop recommendations. Crops taken into consideration are millets, tomato, sugarcane, strawberry, cotton, and rice. Early tests with simulated data have reflected encouraging accuracy. This location-aware, real-time solution can enable farmers to take informed decisions. It also facilitates schemes such as the Soil Health Card Scheme and PM-KISAN. Challenges involve maintaining sensor accuracy and dealing with environmental noise. Periodic calibration will be required for reliable performance. The system decreases reliance on conventional methods and enhances farm profitability. Satellite data and sophisticated ML models could be added in the future. This project provides a cost-effective, scalable step towards smart, sustainable agriculture.

Introduction

Agriculture is a critical sector, providing food, raw materials, and livelihood for millions. However, farmers face challenges such as unpredictable weather, soil erosion, and inefficient resource use. Traditional farming, reliant on experience, often falls short of modern needs. To address this, precision farming using technologies like IoT and Machine Learning (ML) is emerging as a solution for real-time monitoring and decision support.

The project "Crop Prediction Using Sensors and Machine Learning" proposes an intelligent system combining IoT sensors and a Logistic Regression ML model to recommend optimal crops based on environmental data (temperature, humidity, soil moisture, water levels, rainfall). This system aims to improve crop selection, enhance productivity, and optimize resource use while reducing waste.

A literature review highlights the use of satellite data (MODIS) and machine learning models in yield prediction, emphasizing indices like EVI2 and NDWI for accurate crop monitoring. The review also underscores global efforts to reduce hunger and improve food security, noting agriculture's economic importance, especially in India.

The proposed system involves:

• IoT sensor deployment for real-time environmental data collection.

• Data preprocessing to clean and organize sensor data.

• Machine learning (Logistic Regression) for crop prediction based on real-time and historical data.

• A Flask-based web application for farmers to access live sensor readings and crop suggestions.

• Resource optimization to conserve water, manage fertilizers, and increase yield sustainably.

Advantages include real-time forecasting, resource savings, increased productivity, scalability, cost-effectiveness, and user-friendliness. The implementation follows a stepwise approach: data collection, preprocessing, ML model training, web deployment, and continuous system optimization.

Conclusion

In a time when climate change, resource deficiency, and demographic growth present greater threats to food security worldwide, the incorporation of Internet of Things (IoT) and Machine Learning (ML) technologies in agriculture presents an essential and game-changing move ahead. This project, \"Crop Prediction Using Sensors and Machine Learning,\" aims to provide a solution that integrates real-time environmental sensing with predictive analysis in order to enhance crop selection choice, particularly for farmers who belong to under-resourced or rural communities. Through measurement of environmental factors like temperature, humidity, soil water content, rainfall, and water level with Arduino-sensors, and inputting these readings into highly trained ML algorithms such as Logistic Regression and Decision Tree classifiers, this system provides farmers with highly precise, real-time information about which crop is most suitable for their field under prevailing climatic and soil conditions. This model addresses the age-old problem of farmers depending too much on conventional practices or intuition when deciding what to plant. Through the transition from intuition-driven to data-driven farming, the model greatly enhances grassroots-level decision-making. The fact that the solution can suggest crops such as cotton, tomato, sugarcane, strawberry, rice, and millets with dynamic environmental inputs makes the solution both pertinent and scalable. The performance evaluation indicates that ensemble method employing a Voting Classifier achieves higher accuracy, recall, and resilience over isolated models. Reducing the cases of misclassification and efficient imbalanced data processing, the system enables long-term sustainable farming techniques. Moreover, the model reduces wastage of water, minimizes use of fertilizers in excess, and helps make wise land use schemes. Going beyond productivity augmentation alone, the technology provides more prosperous economic returns for farmers, improving their ability to withstand unpredictability in weather patterns. It also complements government initiatives like the Soil Health Card Scheme and PM-KISAN, bringing together innovation and policy in tandem. With minimal hardware and infrastructure investment, the model is an economical and highly effective option compared to capital-intensive precision agriculture equipment, allowing for potential deployment at large scale in different rural and semi-urban areas.

References

[1] Bolton, D. K., and Friedl, M. A. 2013. Forecasting crop yield using remotely sensed vegetation indices and crop phenology metrics. Agricultural and Forest Meteorology 173:74–84. [2] Boryan, C.; Yang, Z.; Mueller, R.; and Craig, M. 2011. Monitoring us agriculture: the us department of agriculture, national agricultural statistics service, cropland data layer program. [3] Geocarto International 26(5):341–358. Boulton, C. A.; Shotton, H.; and Williams, H. T. 2016. Using [4] social media to detect and locate wildfires. In Tenth International AAAI Conference on Web and Social Media. [5] Breiman, L. 2001. Statistical modeling: The two cultures (with comments and a rejoinder by the author). Statist. Sci. 16(3):199–231. [6] Chu, W., and Ghahramani, Z. 2005. Preference learning with gaussian processes. In Proceedings of the 22nd international conference on Machine learning, 137–144. ACM. [7] Cressie, N., and Hawkins, D. M. 1980. Robust estimation of the variogram: I. Journal of the International Association for Mathematical Geology 12(2):115–125. [8] DAAC, N. L. 2015. The modis land products. http://lpdaac.usgs.gov. Dodds, F., and Bartram, J. 2016. The Water, Food, Energy and Climate Nexus: Challenges and an Agenda for Action. Routledge. [9] Ermon, S.; Xue, Y.; Toth, R.; Dilkina, B.; Bernstein, R.; Damoulas, T.; Clark, P.; DeGloria, S.; Mude, A.; Barrett, C.; and Gomes, C. 2015. Learning large-scale dynamic discrete choice models of spatio-temporal preferences with application to migratory pastoralism in East Africa. In AAAI Conference on Artificial Intelligence. [10] FAO. 2015. The state of food insecurity in the world. Meeting the 2015 international hunger targets: Taking stock of uneven progress. [11] Fink, D.; Damoulas, T.; and Dave, J. 2013. Adaptive spatiotemporal exploratory models: Hemisphere-wide species distributions from massively crowdsourced ebird data. In AAAI.

Copyright

Copyright © 2025 D. Sai Kiran, G. Pranay Raj, J. Sai Krishna, K. R. Abhinav, Dr. Obulesu . This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.