Abstract: Attribute recognition model based on entropy weight was used to evaluate the soil fertility of typical region in Sanjiang Plain, which contained 100 soil samples with 7 physical and chemical indexes, including soil total nitrogen, soil organic matter, pH value etc. Then spatial distribution map of soil fertility was delineated on the basis of GIS techniques. Results indicated that the overall level of soil fertility in typical area was higher, 90% of the soil was in a fertile or relative fertile level, soil nitrogen and organic matter had degressive trend. Soil phosphorus and potassium was lower and had extremely unbalanced distributions. Therefore maintaining and advancing soil fertility and actualizing balance fertilization techniques were necessary measures for the region. Evaluation methods in the study had precise recognition function, simple and steady structure and minished the affection of subjective factors to a great extent, which gave information system of soil fertility and expert decision system a powerful technological support. Evaluation results were intuitive and practical, and can provide immediate scientific basis for soil improvement and agricultural industry adjustment.