A Machine Learning Model to Predict Missing People Status

The global problem of the disappearance of people involves many factors, including security, management of public resources and the emotional aspect related to the loss of a loved one. This paper introduces a model that predicts and classifies the status of missing persons, using 20 variables relating to the personal and geographical information of the event. Data was taken from the annual report on disappeared people, published by the Technical Investigation Body of the National Prosecutor's Office of Colombia in 2017, which included 6202 cases. We first reviewed scientific literature associated with machine learning models used to model social phenomena, identifying the most frequently used techniques in these studies. Secondly, the database was debugged in order to proceed with a relational analysis of the variables. Thirdly, three models of supervised data learning were implemented, including decision trees, k-nearest neighbours and random forest. The results show that the random forest model performs consistently better than the other models over the cross-validation and testing stages.