Deep Neuro Fuzzy Model for Crop Yield Prediction

The cornerstone of human civilization, agriculture is essential to social advancement, financial viability, and food security. However, for efficient management, issues like soil health variability and climate change require sophisticated instruments. This study integrates deep neural networks (DNNs) using a fuzzy layer to improve agricultural decision-making in a novel way. The imprecision and unpredictability inherent in agricultural data can pose a challenge for traditional DNNs. In order to solve this, we include a fuzzy phase that uses fuzzy rules to convert crisp inputs into sets of fuzzy values. By processing intricate correlations between variables, this hybrid model enhances the network's capacity to manage ambiguous and noisy data. Despite accuracy around 0.95, traditional DNNs perform well, but they frequently have trouble handling the uncertainty in agricultural data. With an accuracy of 0.96, Convolutional Neural Networks (CNNs) marginally surpass DNNs, especially when it comes to yield forecasting and pesticide recommendation. Nevertheless, with an accuracy of 0.97, the DNN model with a fuzzy layer performs best overall. Our model performs exceptionally well for predicting crop categories, forecasting yields, and suggesting fertilizers and pesticides when inputs like type of crop, rainfall, and area are used. The fuzzy-integrated DNN performs noticeably better than conventional DNNs along with different machine learning models, with an accuracy of 0.97. Fuzzy rules also improve interpretability, making it easier for farmers and agricultural specialists to comprehend the reasoning behind suggestions. This approach is a useful tool for improving crop cultivation and input use since it offers higher prediction accuracy, resilience, and transparency.