Cutting-edge Data Science Techniques

Cutting-edge Data Science Techniques

by Jia Wang, Minseok Song

Cutting-edge technologies are critical in the law enforcement area. Police should have more advanced techniques and tools than criminals have to protect innocent civilians. Since the authorities well understand this fact, they make efforts to use recent technologies. For example, the U.S. law enforcement recently adopted technologies such as GPS tagger, handheld narcotics analyzer, body camera and thermal imaging (Neese, 2015). There are more recent technologies, but we learnt that data science is the centerpiece of them. In this blog, we will first look at cutting-edge data science technologies and their current challenges.

Technology in policing control

(Image from: http://www.bigdatacloud.com/predictive-policing-the-future-of-law-enforcement/)

Crime Predictive Analytical Software

To prevent crimes, police should be able to know criminals’ patterns and predict their next moves. Predictive analytical software can give these powers to the police. An article from Law Enforcement Publications and Conferences reveals that Memphis Police Department (MPD) used IBM SPCC analytics software to reduce violent crime cases and achieved a 29 percent decrease (Phelps, 2010). In the article, John F. Williams, a crime analysis unit manager with MPD, commented that, “We were able to pinpoint when we needed to place our officers in those areas. We saw a tremendous decrease in those areas. From that, we knew we had latched onto something successful.”


Machine Learning for Anti-money Laundering

Machine learning has been being actively used to track anti-money laundering (AML) crimes. A report, Big Shifts What’s Next in AML by Booz Allen Hamilton, highlighted that the next big shift in the fight against financial crime is advanced machine learning (Hamilton, 2014). This report introduced machine learning algorithms including unsupervised similarity clustering, supervised training based on validated true positives and true negatives and binary classification. Data scientists applied these techniques to hundreds of thousands of money transactions to build a predictive model. This model helped them to identify illegitimate traffics and catch criminals.


Image Processing for Identification

Since 2010, the Stockton California Police Department has been using a state-of-the-art device called Mobile Biometric Device (Anderson, 2014). This hand-held device is capable of recognizing suspect’s identification not only by fingerprints but also by facial photos. Once a police takes a photo or scans a fingerprint, this device sends the image to the central database. Then, the database starts analyzing the incoming data with image processing and machine learning techniques. In the article, Erin Mettler, the Stockton Police Department’s fiscal planning and research manager said that, “With the traditional method, you would have to dust the print, lift it with lifting tape, take it back to the office… It could take a day. It could be a week later, depending on the type of crime.” After adopting technology, they could reduce the process to 20 minutes in maximum.


Current Challenges & Limitations

These breakthrough technologies, however, have some limitations. In an article that analyzed problems of data science in law enforcement (Moraff, 2014), the author pointed out that, since predictive analysis is heavily dependent upon data from individual police agencies or software providers, the quality of data can be unreliable. And data quality and content can be inconsistent from city to city, which can decrease analysis accuracy.

No privacy under policing data science

(Image from: https://medium.com/homeland-security/no-privacy-no-security-6fd25bc238c8#.wqf36fu1u/)

In addition, the police databases store not only criminals’ information, but also innocent individuals’ private data. Beth Pearsal at National Institute of Justice in U.S., in his article (Pearsall, 2010) stressed that law enforcement should transparently use these information and form a social consensus in their communities. Communities should have confidence that law enforcement use their data in a right way.


References

Hamilton, B. A. (2014, 1 1). BIG SHIFTS WHAT’S NEXT IN AML. Retrieved 6 3, 2016, from BIG SHIFTS WHAT’S NEXT IN AML: https://www.boozallen.com/content/dam/boozallen/documents/2015/09/machine-learning-and-data-science.pdf

Moraff, C. (2014, 12 15). The Problem With Some of the Most Powerful Numbers in Modern Policing. Retrieved 6 3, 2016, from NEXTCITY: https://nextcity.org/daily/entry/predictive-policing-crime-stats-data-measure

Neese, B. (2015, 3 27). Saving Lives with Technology: Cutting-Edge Law Enforcement Tools. Retrieved 6 3, 2016, from SOUTHERUNIVERSITY: http://online.seu.edu/law-enforcement-tools/

Pearsall, B. (2010, 6 1). Predictive Policing: The Future of Law Enforcement? Retrieved 6 3, 2016, from Office of Justice Program: http://www.nij.gov/journals/266/pages/predictive.aspx

Phelps, C. (2010, 12 1). IBM predictive analytics help slash crime rates in Memphis. Retrieved 6 3, 2016, from Law Enforcement Publications and Conferences : http://www.hendonpub.com/resources/article_archive/results/details?id=1530