Abstract:

The article presents selected decision-making problems concerning the actions of the Police in search of missing persons and proposes methods of machine learning (ML) to support decisions concerning the level of service activities (so-called search level). The problem of determining the scale of operations in searching for missing persons is extremely important, because the health and life of a missing person (often children or elderly people with illnesses) depends on the effectiveness of the search. In this issue, the time from the moment of disappearance to the moment of making a decision should be as short as possible. Equally, decisions taken with limited knowledge, should be appropriate to the situation and risk. A model of the decision-making situation is presented, including a description of information that is processed at all stages of proceedings. Due to the size of the detailed data vector, aggregation functions were proposed, which reduce the size of input data in ML methods. The analysis of historical data allowed to develop and calibrate several ML methods. Moreover, on the basis of formal documents (government act and other regulations), a decision tree was developed for cases which are qualified according to formal regulations. The article contains results of the tests performed on a set of anonymised data from the reality and synthetic data generated to increase the volume of information. In order to identify the most appropriate method of classification, the quality of classification and decision support was studied, which was then presented in the conclusions.