Unsupervised algorithm to classify immigration risk levels

Valles-Coral M.; Lazo-Bartra U.; Pinedo L.; Navarro-Cabrera J.R.; Salazar-Ramírez L.; Ruiz-Saavedra F.; Vidaurre-Rojas P.; Ramírez S. · Data and Metadata (2023) · DOI: 10.56294/dm202398

Resumen

Introduction: Migration is a social phenomenon that affects the structure and distribution of the population, driven by the search for better opportunities and living conditions. In this regard, irregular migration poses a challenge for host countries, as it involves the entry of individuals without the appropriate documentation, potentially compromising national security and border control. Objective: To evaluate the application of the unsupervised DBSCAN algorithm to classify foreigners based on the level of risk of irregular immigration at the National Migration Superintendence of Peru. Methods: We use the DBSCAN algorithm on a dataset from the National Immigration Superintendence, classifying foreigners into clusters according to their level of risk of irregular immigration. In addition, we use the Silhouette, Davies-Bouldin, and Calinski-Harabasz coefficients to evaluate the quality of the classification. Results: DBSCAN classified foreigners into four clusters based on the level of risk of irregular immigration: high, medium-high, medium-low, and low. The performance of the Silhouette index was 0.5338, the Davies-Bouldin index was 0.6213, and the Calinski-Harabasz index was 3680.2359. Conclusions: We show that the use of DBSCAN in the National Immigration Superintendence effectively classified foreigners according to the level of risk of irregular immigration. This tool supports informed decisions of immigration inspectors, favoring Peruvian immigration regulation.

Instituciones: Universidad Nacional de San Martín