An explainable hybrid fuzzy-machine learning framework for multi-crop suitability classification using agro-environmental data

Precise crop type classification and agricultural suitability analysis are critical for optimizing land use, improving productivity, and ensuring food security under climate variability. Traditional models often struggle with overlapping class boundaries, high-dimensional input spaces, and lack of interpretability, limiting their practical deployment in decision-support systems. This study proposes a hybrid fuzzy–machine learning framework that integrates Principal Component Analysis for dimensionality reduction, Fuzzy C-Means for uncertainty-aware clustering, and a Random Forest classifier for robust prediction. The framework was trained and validated on a real-world dataset of soil and climatic parameters (soil pH, Nitrogen, Phosphorus, Potassium, temperature, humidity, and wind speed) covering seven major crops, with class-wise stratified five-fold cross-validation ensuring reliability. Principal Component Analysis reduced redundancy while retaining over 87 % of total variance, and Fuzzy C-Means membership values captured uncertainty in overlapping crop suitability patterns. The combined Principal Component Analysis, Fuzzy C-Means clustering, and Random Forest model achieved an average accuracy of 93.8 %, precision of 94.3 %, recall of 94.1 %, and F1-score of 94.1 %, representing a 6 % gain in F1-score over baseline Random Forest. Crops such as Cotton, Sugarcane, and Soybean consistently showed high classification performance, while errors were concentrated in pairs with overlapping requirements (Potato vs. Soybean and Sugarcane vs. Rice). SHapley Additive Explanations-based analysis revealed that The first principal component and fuzzy membership features contributed most to decision-making, aligning with agronomic knowledge and providing interpretability. These results confirm that the proposed framework improves accuracy, enhances transparency, and offers practical decision-support value for sustainable agricultural planning.