Integrated approach to personal identification using dermatoglyphs and artificial neural networks

Abstract

Introduction. Due to the full-scale military invasion of Ukraine by the aggressor state, the tendency to aggravate local armed conflicts in the world, which causes a large number of depersonalised, fragmented corpses, the problem of identifying the bodies of two or more persons arises

The aim of the study to develop expert criteria for the informativeness of dermatoglyphic fingerprints in the system of forensic medical identification of a person.

Materials and methods. The object of the study was fingerprint cards obtained from 460 people (200 women and 260 men) aged 18-59 years living in Ukraine. We used statistical analysis and neural network programming.

Results. Using neural network prediction, we have developed a methodology for reproducing unknown (lost) phenotypic traits based on the available ones (dermatoglyphs). Given the fact that many different neural networks can be built even on the same variables, depending on their combination, we managed to achieve a prediction accuracy of 73-90%, which suggests that a combination of different neural networks and an integrated approach show better results.

Conclusions. Based on the above, it can be concluded that the reliability of the results obtained ranged from 73-90% (automatically calculated by the Dermatoglyphics For Prediction (DFP) software), which is significantly higher than the results of previous fingerprint examinations. The use of our proposed software in combination with other basic methods will improve the quality of forensic identification examinations.