Soil fertility grading with Bayesian Network transfer learning

Soil fertility grading is an important issue in the agriculture domain, AI based approach has been applied recently. But in most circumstance data obtaining is a expensive and time consuming procedure, sometimes even impossible. This paper presents a Bayesian Network based transfer learning algorithm. The existing training results can be transferred between the nearby land squares. The proposed algorithm considers both the similarity between the learning task and the geographical position of the land squares. Empirical experiment were implemented to prove the efficiency of the algorithm.