Related Experiment Videos
A data-driven methodology to discover similarities between cocaine samples
Fidelia Cascini1, Nadia De Giovanni2, Ilaria Inserra3
1Department of Life Sciences and Public Health, Università Cattolica del Sacro Cuore, 00168, Rome, Italy. fidelia.cascini1@unicatt.it.
Scientific Reports
|September 30, 2020
Summary
This study introduces a machine learning system for classifying cocaine profiles, aiding law enforcement. The system analyzes chemical data to identify drug origins and preparation methods with high accuracy.
Area of Science:
- Forensic Science
- Data Science
- Computational Chemistry
Background:
- Machine learning applications are widespread in science but have not been applied to illegal drug analysis.
- Chemical profiling generates extensive data crucial for understanding drug trafficking patterns.
Purpose of the Study:
- To develop a novel web-based system for cocaine classification and comparison using machine learning.
- To create a tool that aids intelligence actions by standardizing drug profile analysis and identifying sample relationships.
Main Methods:
- Development of the Profiling Relations In Drug trafficking in Europe (PRIDE) system, a web-based platform.
- Application of machine learning algorithms for classifying and comparing cocaine chemical profiles.
- Evaluation of algorithms using precision, recall, and F0.5-measure against a gold standard.
Main Results:
- The system achieved 91% precision, 78% recall, and an 88% F0.5-measure.
- Demonstrated the capability of machine learning to automatically classify cocaine profiles.
- Highlighted the system's effectiveness in determining the probable common origin, batch, or preparation process of drug samples.
Conclusions:
- Machine learning is a viable and effective tool for the automatic classification of cocaine profiles.
- The PRIDE system offers a standardized and powerful methodology for forensic intelligence across Europe.
- This approach can significantly enhance investigations into drug trafficking networks.