The precision agriculture domain has become highly dependent on sophisticated data analytics techniques for achieving accurate prediction of crops' yield and monitoring of the soil. However, dealing with heterogeneous agricultural data has been recognized as one of the most difficult problems for AI researchers. Therefore, this research proposes a combined approach based on the integration of Hypergraph Neural Network (HGNN), Liquid Neural Network (LNN), and Deep Evidential Learning (EDL). Specifically, HGNN is able to capture complex relations between different features of soils' NPK composition, climate, crops, and other parameters that are field-level specific. Meanwhile, LNN allows modeling continuous streams of sensors' data as they describe changing conditions of the environment. Finally, to increase decision-making reliability, EDL is utilized for yield predictions with the quantified degree of uncertainty for the recommendations provided to farmers and agronomists. Overall accuracy of the solution equals to 98.30%. This model is capable of showing how relational modeling, together with adaptive temporal learning and evidential reasoning can revolutionize the field of agricultural analytics. With this combination of these various methods, the model not only succeeds in being highly accurate in its predictions but also manages to provide confidence-aware information that is essential for making decisions in the case of uncertainty. This way of analysis allows farmers to manage their resources effectively, predict yields, and deal with any changes in soil and climate.
