Nutritional status of children is significant for the child, which are the major predicator to achieve good health. The deficiency in nutrition affects the children health. Therefore, the lack of adequate nutritional status in children has been a serious problem which brings under nutrition for children, that widespread in Ethiopia which affects the children in different areas, including both the rural and urban areas. The diagnosis and treatment of children is challenged who are affected with under nutrition and increase death and disease because, to know children dis order is very difficult and availability of physicians are less to treat children which means there is scarcity of physicians as well as health facilities to treat the children on time. Therefore, to solve this problem, the main objective of this research is to develop the knowledge based system to identify the under nutritional status of children through diagnose and treatment for the under nutrition children, in order to support the physicians by facilitating the diagnosis and their treatments. To achieve this objective, knowledge is acquired using both structured and unstructured interviews with six experts, which are selected purposely from Felege Hiwot Referral Hospital (FHRH). In addition to that, knowledge is acquired from secondary sources (internet, articles, manuals and some reports). The acquired knowledge was modeled using decision tree to represent with concepts of sign, symptom, risk factors, physical and laboratory tests involved in under nutrition diagnoses purpose. The study used rule based reasoning method through backward reasoning approach for diagnoses under nutrition of children below five years, and the prototype was developed with SWI Prolog. The main significance of this study is to diagnose and provide any treatments of under nutrition children for the purpose increasing the quality of many decision-making tasks for researchers, the health sector and other stakeholders to facilitate a case of nutrition in children. In addition to this, it reduces costs and time spent for decision-making with the viii manual working of diagnosis system by reducing the burden of tasks to human experts. Generally, the performance of the system can be evaluated in two ways of evaluation; those are system performance testing using test cases that has 85% and user acceptance testing with 86.2% that have a good performance to achieve its objective. Therefore it has the overall performance is 85.7%.