1. Abstract

Explainable Crop-Yield Prediction under Climate Variability is a machine learning project designed to predict agricultural crop yields by analyzing weather conditions, soil characteristics, and historical crop production data. Climate variability, including changes in temperature, rainfall, humidity, and drought conditions, can significantly affect agricultural productivity. Traditional crop-yield prediction methods may struggle to capture complex relationships between environmental factors and crop growth. This project uses machine learning algorithms such as Random Forest, XGBoost, and Linear Regression to estimate crop yields under different climatic conditions. Explainable AI techniques such as SHAP (SHapley Additive exPlanations) are integrated to identify how individual environmental and agricultural features influence predictions. The system provides yield estimates, feature-importance visualizations, and explanations of climate-related impacts through an interactive dashboard. The proposed solution aims to help farmers, agricultural researchers, and policymakers understand yield variations, improve crop planning, and make informed decisions under changing climate conditions.

2. Objectives

1. To develop a machine learning model for predicting crop yields under varying climatic conditions.
2. To analyze the influence of temperature, rainfall, humidity, and soil properties on crop productivity.
3. To integrate explainable AI techniques to interpret crop-yield predictions.
4. To identify the most influential environmental and agricultural factors affecting yields.
5. To compare machine learning algorithms based on prediction performance.
6. To visualize crop-yield trends and climate-related impacts through an interactive dashboard.
7. To support data-driven agricultural planning and climate adaptation strategies.

3. Existing System

Existing crop-yield estimation methods commonly rely on historical production statistics, traditional agricultural knowledge, statistical models, and manual analysis of weather and soil conditions. Some systems use machine learning to predict yields but provide limited explanations of how environmental factors influence their results.

Disadvantages of the Existing System

1. Traditional methods may not capture complex climate-yield relationships.
2. Manual analysis of agricultural and weather data can be time-consuming.
3. Sudden changes in rainfall and temperature may reduce prediction reliability.
4. Some prediction models operate as black boxes.
5. Farmers may find it difficult to identify the factors responsible for predicted yield changes.
6. Limited integration of climate data may reduce the usefulness of yield forecasts.

4. Proposed System

The proposed system combines machine learning and Explainable AI (XAI) to predict crop yields and explain the influence of climate variability on agricultural productivity.

The system collects historical crop-yield records, weather observations, and soil information. The collected data is cleaned, processed, and used to train models such as Random Forest, XGBoost, and Linear Regression. The selected model predicts expected crop yield based on the available input conditions. SHAP is then used to explain the contribution of features such as rainfall, temperature, soil properties, and humidity to the prediction.

An interactive dashboard presents predicted yields, climate trends, feature importance, and individual prediction explanations. By understanding the factors influencing predicted productivity, users can better evaluate potential agricultural risks and explore suitable adaptation strategies. Prediction quality depends on the quality, coverage, and regional relevance of the training data.

5. Implementation Procedure

1. Dataset Collection: Collect historical crop-yield data, climate records, and soil information from Kaggle or reliable agricultural and meteorological sources.
2. Data Preprocessing: Handle missing values, remove duplicate records, standardize date formats, and encode categorical variables.
3. Data Integration: Combine crop production, rainfall, temperature, humidity, soil type, and other relevant information by location and growing season.
4. Exploratory Data Analysis: Visualize climate patterns, crop-yield distributions, seasonal changes, and relationships between environmental factors and productivity.
5. Feature Engineering: Create useful features such as seasonal rainfall totals, temperature averages, rainfall anomalies, and drought indicators where data is available.
6. Model Training: Train machine learning algorithms such as Random Forest, XGBoost, and Linear Regression to estimate crop yield.
7. Model Evaluation: Evaluate performance using MAE, RMSE, and \(R^2\) score. Use time-based or location-based validation where appropriate to reduce data leakage.
8. Explainable AI Integration: Apply SHAP to explain global feature importance and individual crop-yield predictions.
9. Dashboard Development: Build a Streamlit dashboard displaying predicted yield, climate conditions, historical trends, and SHAP explanation plots.
10. Testing and Deployment: Test the system with different crops, regions, and climate scenarios, then deploy the application locally or on a suitable hosting platform.

6. Software Requirements

7. Hardware Requirements

Recommended configuration: Intel Core i5 or equivalent, 8 GB RAM, and SSD storage.

8. Advantages of the Project

1. Improved yield prediction: Estimates crop productivity using historical agricultural and climate data.
2. Climate impact analysis: Helps identify how rainfall, temperature, and other weather variables influence yield.
3. Explainable predictions: Uses SHAP to show which factors contribute to individual predictions.
4. Data-driven crop planning: Supports agricultural planning based on available environmental information.
5. Risk identification: Helps highlight potentially unfavorable climatic conditions.
6. Interactive visualization: Presents yield forecasts, climate trends, and feature importance through charts.
7. Model comparison: Allows evaluation of multiple machine learning algorithms.
8. Research support: Helps agricultural researchers study climate variability and crop productivity.
9. Scalability: Can be extended to additional crops, locations, soil characteristics, and climate indicators.