Jithendra et al. (2025) Two-stage rainfall forecasting and crop classification using puma-optimized ANFIS–SVM

Identification

• Journal: Smart Agricultural Technology
• Year: 2025
• Date: 2025-12-13
• Authors: Thandra Jithendra, A Divya, Ibrahim Aljubayri, T Haritha, Mohammad Zubair Khan
• DOI: 10.1016/j.atech.2025.101719

Research Groups

• Department of Mathematics, Koneru Lakshmaiah Education Foundation, Hyderabad, Telangana, India
• School of Technology, The Apollo University, Chittoor, Andhra Pradesh, India
• Department of Computer Science and Information, Imam Mohammad Ibn Saud Islamic University (IMSIU), Riyadh, Saudi Arabia
• Department of Computer Science and Engineering, Vemu Institute of Technology, Chittoor, Andhra Pradesh, India
• Faculty of Computer and Information System, Islamic University of Madinah, Madinah, Saudi Arabia

Short Summary

This study developed an ANFIS-SVM-PO framework for agricultural recommendations, achieving 99.15% accuracy in rainfall prediction and 90% accuracy in optimal crop classification using an Indian agricultural dataset.

Objective

• To design and evaluate a hybrid Adaptive Neuro-Fuzzy Inference System (ANFIS), Support Vector Machine (SVM), and Puma Optimizer (PO) machine learning framework (ANFIS-SVM-PO) for a two-stage agricultural recommendation system, optimizing rainfall prediction and subsequent optimal crop classification based on ecological factors.

Study Configuration

• Spatial Scale: Agricultural dataset from India.
• Temporal Scale: A dataset comprising 48 samples; no explicit temporal range is specified for the data collection.

Methodology and Data

• Models used: Adaptive Neuro-Fuzzy Inference System (ANFIS), Support Vector Machine (SVM), Puma Optimizer (PO). Comparative models included ANFIS-SVM, ANFIS-SVM-PSO, ANFIS-SVM-COOT, ANFIS-SVM-RSA, and ANFIS-SVM-COA.
• Data sources: Publicly available "Crop recommendation dataset (2020)" from Kaggle (https://www.kaggle.com/datasets/atharvaingle/crop-recommendation-dataset). Input features included nitrogen, phosphorus, potassium content in soil (ratio 0-100), soil pH value (5-10), temperature (degree Celsius), relative humidity (percentage 50-100), and rainfall (millimeters 40-300).

Main Results

• The ANFIS-SVM-PO framework demonstrated superior performance compared to other models.
• For rainfall prediction, ANFIS-SVM-PO achieved:
  • Root Mean Squared Error (RMSE): 6.9312 mm
  • Mean Absolute Error (MAE): 5.7083 mm
  • Mean Absolute Percentage Error (MAPE): 0.0595
  • Coefficient of Determination (R²): 0.9915
  • Accuracy: 99.15%
• For optimal crop classification (rice, maize, watermelon, coffee), ANFIS-SVM-PO achieved:
  • Accuracy: 90%
  • Precision: 88%
  • Recall: 87%
  • F1-Score: 78%

Contributions

• Presentation of an innovative hybrid machine learning framework (ANFIS-SVM-PO) for agricultural recommendation mechanisms.
• Improvement of ANFIS and SVM model performance by integrating the Puma optimizer to address overfitting, local optima, and achieve faster convergence rates.
• Development of a two-stage framework where ANFIS-PO initially predicts rainfall using input features, and SVM-PO subsequently selects the optimal crop based on predicted rainfall and soil characteristics.
• Empirical validation highlighting the superior efficiency of ANFIS-SVM-PO compared to ANFIS-SVM, ANFIS-SVM-PSO, ANFIS-SVM-COOT, ANFIS-SVM-RSA, and ANFIS-SVM-COA.

Funding

• Research grant funded by the Deanship of Scientific Research, Islamic University of Madinah, Madinah, Saudi Arabia.

Citation

@article{Jithendra2025Twostage,
  author = {Jithendra, Thandra and Divya, A and Aljubayri, Ibrahim and Haritha, T and Khan, Mohammad Zubair},
  title = {Two-stage rainfall forecasting and crop classification using puma-optimized ANFIS–SVM},
  journal = {Smart Agricultural Technology},
  year = {2025},
  doi = {10.1016/j.atech.2025.101719},
  url = {https://doi.org/10.1016/j.atech.2025.101719}
}