Artificial Intelligence in Agriculture: A Systematic Review of Crop Yield Prediction and Optimization

Aim:

           To develop a robust and accurate crop yield prediction system by integrating meteorological data, pesticide usage records, and crop yield statistics, leveraging advanced machine learning techniques to promote sustainable agricultural practices and enhance global food security.

Abstract:

      Climate change and excessive pesticide use pose significant challenges to agricultural productivity and global food security. Accurate crop yield prediction is critical for addressing these issues and promoting sustainable agricultural practices. This study introduces a machine learning-based framework that integrates meteorological data, pesticide usage records, and crop yield statistics to forecast crop yields effectively.

       The research focuses on data preprocessing, model development, and evaluation using machine learning techniques. By analyzing the relationships between environmental factors and crop yields, the study provides insights into optimal agricultural conditions. The findings emphasize the role of data-driven methods in improving resource management, supporting sustainable farming, and enhancing resilience to climate change. This work contributes to advancing predictive tools for agriculture and ensuring long-term food security. A dedicated AI Agent is incorporated into the framework, operating with a Prediction Mode that delivers accurate yield forecasts from processed meteorological and field data. Its Suggestion Mode converts those predictions into recommendations, guiding farmers toward better resource allocation and adaptive management strategies. This dual-mode agent strengthens data-driven agricultural decision-making and supports climate-resilient farming practices. Overall, the work advances predictive intelligence for agriculture and contributes to long-term food security.

Proposed System:

          The proposed system enhances crop yield prediction by integrating meteorological data, pesticide usage records, and historical yield information into a unified processing pipeline. After ensuring data quality through preprocessing, machine learning models such as KNeighborsRegressor, RandomForestRegressor, DecisionTreeRegressor, and BaggingRegressor are trained and optimized using GridSearchCV. An AI Agent is incorporated into the framework, with a Prediction Mode driven by the trained models and a Suggestion Mode powered by Llama-3.3-70B-Versatile to generate context-aware recommendations for agricultural decisions. This dual-mode agent is deployed through a Chainlit-based web application, providing an interactive prediction and advisory interface for end users. The system streamlines yield forecasting, improves decision-making, and supports data-driven agricultural management.

Advantage:

• Agent AI integrates meteorological data, pesticide usage records, and historical yield information, creating a unified and holistic understanding of the factors that influence crop productivity.
• Agent AI applies machine learning models and optimizes them with GridSearchCV, delivering reliable and accurate crop yield predictions that support informed decisions for farmers and agricultural planners.
• Agent AI promotes sustainable agricultural practices by generating data-driven insights that help optimize pesticide usage and overall resource management.
• Agent AI is deployed through an intuitive Chainlit-based web application, enabling users to easily access predictions and interact with the system in real time.