Artificial Intelligence Driven Crop Price and Yield Prediction Using ML for Sustainable Agriculture

Abstract

This work suggests a machine learning based approach to improve agricultural decision-making by utilising real-time data to increase yield accuracy, lower climate-related risks, and support sustainable farming practices, which uses Polynomial Regression (PR) and Random Forest Algorithm (RFA) to forecast crop production in India. Materials and methods. In order to examine crop production, crop recommendation and price prediction patterns in India, the materials and methodologies employed in the dataset from the given Crop prediction URL, crop recommendation URL and fertilizer prediction URL include the use of multivariate datasets that comprise numerical data and categorical data. In order to train machine learning models, the dataset is pre-processed by handling missing values, encoding categorical variables, and normalising or scaling numerical features. Based on these many characteristics, models such as RFA and Machine Learning (ML) are commonly employed to forecast crop yield.

Results and discussion. Model performance is evaluated by comparison of predicted values with time real results. Metrics such as MAPE and RMSE are used to assess accuracy. The best-performing model is selected for implementation based on these evaluations. In this study, projected values and real-time outcomes were compared to assess the model's performance. Accuracy was evaluated using metrics including Root Mean Squared Error (RMSE) and Mean Absolute Percentage Error (MAPE). With a MAPE of 8.12% and an RMSE of 24.5, the RFA outperformed the Decision Tree (MAPE: 10.45%, RMSE: 30.2) and PR (MAPE: 9.25%, RMSE: 27.8). These assessments demonstrated the Random Forest model's greater prediction accuracy, which led to its selection for deployment.

Conclusion. Our approach helps farmers make better decisions, lower financial risks, and increase agricultural output by utilizing both historical and current data.  We found major obstacles to pricing and yield prediction, including climate variability and inconsistent data, by conducting a thorough analysis of the body of current work. Farmers may get a greater understanding of market trends, soil health, and climatic patterns by integrating AI-driven agricultural forecasts with village-level development initiatives.