The case-base reasoning to support decision-making in humanitarian response

Abstract

The object of the study is decision-making processes in humanitarian response. It is noted that certain experience of humanitarian response can be used in similar situations in the future. It is determined that humanitarian response processes have the following features: complexity and dynamism; decision-making takes place in conditions of uncertainty; emerging situations require a prompt response. There is a lack of an integrated approach to the creation of decision support systems in humanitarian response. Current research is related to decision-making in the framework of predicting and eliminating the consequences of natural disasters, man-made disasters and armed conflicts, and humanitarian issues are considered as related. Case-base reasoning was chosen as a representation of knowledge about humanitarian response processes. This method allows using previous decision-making experience without performing a detailed analysis of the subject area and adapting the accumulated knowledge to new situations. The functions and structure of an decision-making support system in humanitarian response are defined. The system uses the case-base reasoning module to represent knowledge, which simulates human reasoning and is based on the effective use of existing experience. The module implements a modified case-base reasoning. The parametric representation of a case is extended by the indicator of the use of precedents. The structured representation of the case solution in the form of a sequence of steps, each of which has its own characteristics, allows for the adaptation of several cases that are close to the current situation. The results of the study were experimentally verified by training the module on the initial data set. Fifty cases were generated for the experiment, describing decision-making in the provision of drinking and industrial water to the population. The training results show that when the cases database is filled with fifty cases, the classification quality is 89%. The results obtained allow us to conclude that it is expedient to structure knowledge about humanitarian response in the form of cases with the possibility of further adaptation to new situations and training.