CEPOL Research & Science Conference 2022 MRU, Vilnius

Over the past few years it has been noted that decisions in the scope of public policy, including those focused on citizens’ security as well as the prevention of crime, should be based on scientific evidence. All of that with the objective of avoiding a decision making process substantiated by biased data. One of the security fields which generate more public concern is the missing persons phenomenon. In Spain, since 2018, the National Centre for Missing Persons (CNDES), dependent from the Spanish Ministry of Interior, has been promoting research in collaboration with Academia to improve instruments and investigation tools for missing person cases. Specifically, a predictive risk assessment system (“SER – DesVi”) has been developed to identify those missing person cases which result in harm, suicide or homicide outcomes. All of that is based on a checklist formed by empirical risk and protective factors which will also serve as a standardized tool for collecting historical data. A proportionate stratified sampling was conducted. In this manner, a representative sample of missing person cases was collected. For the construction of the predictive scales for harm, suicide and homicide, a relational analytical study of cases and controls was designed and bivariate (Chi-Squared) and multivariate (Binary Logistic Regression) statistical techniques were employed. Having been constructed, the scales were evaluated for validity and predictive capability using the test discrimination estimators (odd ratios – OR, area under the curve –AUC, sensitivity and specificity) and calibration estimators (positive predictive value – PPV and negative predictive value – NPV). Finally, different machine learning algorithms (Logic Rules, Decision Trees, Random Forest and Bayesian Networks) were explored to facilitate the automatic calibration of the indicators which compose the predictive scales. In the study of missing persons, the development of algorithmic and predictive models has enabled the construction and calibration of the “SER – DesVi” system which entail implications for a) the prioritisation of investigation lines, b) the adoption of a more effective strategy for the management of police investigation resources, c) the investigation of cold cases and d) the establishment of an evidence-based approach in the criminal justice system.