Intelligent Crop Recommendation System Utilizing Fuzzy Logic and Bayesian Network Approaches for Dynamic Environmental Adaptation

Accurate crop selection plays a crucial role in improving agricultural productivity, yet traditional decision-making methods often struggle to handle uncertainty arising from varying soil nutrients and climatic conditions. This study aims to develop an intelligent crop recommendation system using a hybrid Artificial Intelligence approach that integrates Fuzzy Logic and a Tree Augmented Naive Bayes (TAN) model. Fuzzy Logic is employed to process continuous and uncertain environmental variables, while the TAN model is used to perform probabilistic crop classification by modeling dependencies among features. The research methodology includes data preprocessing, fuzzification of soil and climate parameters, probabilistic structure learning using the TAN algorithm, and inference for crop recommendation. System implementation and testing were conducted using Python-based simulations with scikit-fuzzy and pgmpy libraries across multiple experimental scenarios. The system is evaluated using the Crop Recommendation Dataset, created by Atharva Ingle and publicly available on Kaggle at https://www.kaggle.com/datasets/atharvaingle/crop-recommendation-dataset. The results demonstrate that the proposed system effectively captures non-linear relationships and uncertainty in agricultural data, achieving an accuracy of 97.73% in recommending 22 different crop varieties, which shows that the hybrid Fuzzy–TAN approach provides a reliable and accurate decision support system for crop selection, offering improved adaptability and performance compared to conventional recommendation methods.精准的作物选种对于提升农业生产力至关重要，但传统决策方法往往难以有效应对因土壤养分波动与气候条件变化带来的不确定性。本研究旨在开发一种融合模糊逻辑（Fuzzy Logic）与树增强朴素贝叶斯（Tree Augmented Naive Bayes，TAN）模型的混合人工智能方法，以构建智能作物推荐系统。其中，模糊逻辑用于处理连续型与不确定性环境变量，而TAN模型则通过建模特征间的依赖关系，实现概率化作物分类。 本研究的方法论涵盖数据预处理、土壤与气候参数的模糊化处理、基于TAN算法的概率结构学习，以及用于作物推荐的推理环节。系统的实现与测试依托基于Python的仿真环境，借助scikit-fuzzy与pgmpy库开展多组实验场景验证。 本研究所用的评估数据集为Atharva Ingle构建的作物推荐数据集（Crop Recommendation Dataset），该数据集已公开至Kaggle平台，访问链接为https://www.kaggle.com/datasets/atharvaingle/crop-recommendation-dataset。实验结果表明，所提出的系统能够有效捕捉农业数据中的非线性关系与不确定性，在22种不同作物品种的推荐任务中实现了97.73%的准确率，证实该混合模糊-TAN方法可为作物选种提供可靠且精准的决策支持系统，相较于传统推荐方法具备更优异的适应性与综合性能。