XAI-Based SmartAgriGo: An Intelligent Agriculture Framework for Transparent Crop Recommendation and Plant Disease Detection

Abstract

Agriculture in India is challenged due to inappropriate crop selection, climate change, soil nutrient imbalance, and late identification of plant diseases. To overcome these problems, this paper proposes SmartAgriGo, an Explainable Artificial Intelligence (XAI)-based smart agriculture framework for transparent crop recommendation and automated plant dis-ease identification. The proposed framework combines machine learning, and explainable AI for accurate and interpretable agri-cultural decision support. Crop recommendation is done based on soil nutrients, pH, temperature, humidity and rainfall, where XLNet-based feature extraction and Support Vector Machine (SVM) classification identify the best-suited crop. Plant disease identification is done based on Convolutional Neural Network (CNN) and Softmax classification of leaf images. To improve interpretability, SHAP values are used for crop recommendation, and LIME values are used for disease identification.The interface designed for farmers shows the prediction results with confidence and explanation. SmartAgriGo fills the gap between state-of-the-art AI approaches and real-world agriculture by providing accurate, interpretable, and data-driven agricultural support.

Introduction

The text presents SmartAgriGo, an Explainable AI (XAI)-enabled smart agriculture system designed for Indian farming conditions. The system combines artificial intelligence, machine learning, deep learning, and explainability techniques to help farmers make reliable decisions about crop selection and plant disease detection.

Background and Motivation

Agriculture is a major contributor to India's economy, but many farmers still rely on traditional methods based on experience. They face challenges such as:

• Unpredictable weather conditions.
• Incorrect crop selection.
• Soil nutrient imbalance.
• Late detection of plant diseases.

AI-based agriculture systems can improve farming by using data-driven predictions for crop recommendation, irrigation, and disease detection. However, many existing AI systems act as black boxes, giving predictions without explaining their reasoning. They also often depend on unreliable or non-authoritative datasets, reducing farmer trust.

Proposed System: SmartAgriGo

SmartAgriGo is an integrated AI-based agricultural decision-support platform that provides:

1. Crop Recommendation
   • Uses soil and weather parameters such as:
     • Nitrogen, Phosphorus, Potassium (NPK)
     • Soil pH
     • Temperature
     • Humidity
     • Rainfall
   • Uses machine learning models to recommend suitable crops.
2. Plant Disease Detection
   • Uses leaf images uploaded by farmers.
   • Applies deep learning techniques to identify plant diseases and their stages.
3. Explainable AI (XAI)
   • Uses SHAP and LIME techniques to explain AI predictions.
   • Shows which features or image regions influenced the decision.
   • Improves transparency, trust, and acceptance among farmers.

The system uses authentic government-approved agricultural datasets and provides predictions through mobile and web interfaces.

Literature Survey Findings

Crop Recommendation Systems

Previous studies have used:

• Support Vector Machines (SVM)
• Random Forest
• Gradient Boosting
• Other machine learning algorithms

These systems use soil and climate information to predict suitable crops. However, many lack:

• Interpretability.
• Real-time environmental adaptation.
• Reliable government-based datasets.

Plant Disease Detection

CNN-based models have achieved high accuracy in detecting diseases from leaf images. However, most systems:

• Work as black boxes.
• Do not explain disease predictions.
• Use publicly available datasets that may not match regional agricultural conditions.

Explainable AI in Agriculture

XAI methods such as:

• SHAP (SHapley Additive Explanations) — explains the contribution of input features.
• LIME (Local Interpretable Model-Agnostic Explanations) — highlights important regions in images.

These techniques make AI decisions understandable for farmers.

Algorithms Used in SmartAgriGo

1. Support Vector Machine (SVM)

• Used for crop recommendation.
• Classifies crops based on soil and weather conditions.
• Works well with moderate-sized agricultural datasets.

2. Random Forest

• Improves crop prediction accuracy.
• Reduces overfitting.
• Provides feature importance information.

3. CNN with Softmax

• Used for plant disease identification.
• Extracts features from leaf images.
• Classifies diseases and provides confidence scores.

4. SHAP and LIME

• Explain AI outputs.
• SHAP explains crop recommendations.
• LIME explains disease detection by highlighting affected leaf areas.

Practical Challenges

The implementation of SmartAgriGo faces several challenges:

• Limited resources: Rural users may have low-powered devices, requiring optimized AI models.
• Internet limitations: Poor connectivity may affect cloud-based processing.
• Data security: Farmer data must be securely stored and managed.
• Scalability: The system must support different regions, climates, and crops.
• Environmental changes: Continuous model updates are required due to changing weather and farming conditions.

Possible solutions include:

• Model compression and optimization.
• Edge-cloud computing.
• Secure data handling.
• Continuous AI model retraining.

Implementation and Evaluation

The system is implemented using:

• Scikit-learn for SVM and Random Forest models.
• TensorFlow for CNN-based disease detection.

Data preprocessing includes:

• Handling missing values.
• Feature normalization.
• Label encoding.
• Image resizing and augmentation.

Performance is evaluated using:

• Prediction accuracy.
• Precision and recall.
• F1 score.
• Confidence scores.
• Reliability of SHAP and LIME explanations.

Expected Outcome

SmartAgriGo aims to provide farmers with a reliable, transparent, and intelligent agricultural assistant that can:

• Recommend suitable crops based on environmental conditions.
• Detect plant diseases early.
• Explain AI decisions clearly.
• Improve productivity and reduce farming risks.

Conclusion

This research introduced SmartAgriGo, an Explainable Ar-tificial Intelligence (XAI)-based intelligent agriculture frame-work for transparent crop suggestion and plant disease identi-fication [8], [10]. The proposed framework combines super-vised machine learning models for structured soil and cli-matic data analysis, and Convolutional Neural Networks with Softmax activation for multi-class plant disease identification. Unlike traditional black-box agricultural prediction models, SmartAgriGo combines Explainable AI approaches, including SHAP and LIME, to improve transparency and trustworthi-ness. SHAP offers quantitative feature attribution for crop suggestion, and LIME offers visual explanations for disease identification. This combination fills the transparency gap in predictive models, which is essential in farmer-centric systems where predictions have a direct effect on economic outcomes. The probability outputs produced by the Softmax layer further improve the reliability of the decision by providing confidence levels for each classification outcome. In summary, SmartA-griGo provides a unified, scalable, and interpretable solution for smart agriculture. The proposed framework provides a solution for sustainable agriculture practices. The framework uses predictive analytics, deep learning, and explainability to promote the adoption of responsible AI in the agricultural sector.

References

[1] A. Agarwal, S. Ahmad, and A. Pandey, “Crop Recommendation Based on Soil Properties: A Comprehensive Analysis,” in Proc. 14th Int. Conf. Computing Communication and Networking Technologies (ICCCNT), 2023. [2] Ashwitha and C. A. Latha, “Crop Recommendation and Yield Estimation Using Machine Learning,” in Proc. IEEE Int. Conf. Emerging Technolo-gies in Agriculture, pp. 88–94, 2023. [3] S. Bag, N. Padhy, and R. Panigrahi, “Enhancing Crop Recommendation Systems Using Deep Learning Techniques on Soil & Environmental Data,” in Proc. IEEE Int. Conf. Computing Communication and Net-working Technologies, 2025. [4] D. Balakrishnan et al., “Agricultural Crop Recommendation System,” in Proc. 3rd Int. Conf. Intelligent Technologies (CONIT), 2023. [5] J. R., H. D., and P. B., “A Machine Learning-based Approach for Crop Yield Prediction and Fertilizer Recommendation,” in Proc. 6th Int. Conf. Trends in Electronics and Informatics (ICOEI), pp. 1–6, 2022. [6] S. Kumar and M. Kumar, “Developing an XAI-Based Crop Recom-mendation Framework Using Soil Nutrient Profiles and Historical Crop Yields,” IEEE Trans. Consumer Electronics, vol. 71, no. 3, pp. 1204–1215, 2025. [7] J. J. Liu, H. Wu, and I. Riaz, “Advanced Technologies for Smart Fertilizer Management in Agriculture: A Review,” IEEE Access, vol. 10, pp. 45192–45210, 2022. [8] R. J. Martin et al., “XAI-Powered Smart Agriculture Framework for Enhancing Food Productivity and Sustainability,” IEEE Access, vol. 12, pp. 1–12, 2024. [9] N. Patel and R. Deshmukh, “AI Model for Crop and Fertilizer Recom-mendation Using Traditional Soil and Climate Datasets,” IEEE Trans. Fuzzy Systems, vol. 31, no. 4, pp. 2256–2265, 2023. [10] R. Screpnik et al., “Artificial Intelligence in Agriculture: A Systematic Review of Crop Yield Prediction and Optimization,” IEEE Access, vol. 13, pp. 7845–7860, 2025. [11] R. Sharma and M. Tiwari, “Crop Disease Detection using CNNs,” in Proc. IEEE Int. Conf. Machine Learning and Applications (ICMLA), pp. 1021–1027, 2024. [12] V. Tomar et al., “Development of a Machine Learning-Based System for Optimizing Crop Recommendations,” in Proc. IEEE Int. Conf. Advances in Computing, Communication and Networking (ICAC2N), 2024.

Copyright

Copyright © 2026 Dr. Dipalee Rane, Ms. Arusha Gade, Ms. Tanushka Chaudhary, Ms. Bhakti Darawade, Ms. Ashlesha Pal, Mrs. Ashwini Athawale. 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.