Who We Are
We develop predictive policing applications using cutting edge crime prediction algorithms using computational criminology, machine learning, theoretical physics, geospatial technology.
SP is selected as a recipient of a contract research adopted by NICT (National Institute of Information and Communications Technology, Japan) from 2018 and she is responsible for building algorithms and software tools for crime prediction.
Crime Prediction System "CRIME NABI"
Crime Prediction System "CRIME NABI" is an original AI engine developed by Singular Perturbations Inc..
CRIME NABI system requires committed crime data as input, along with other explanatory local variables such as weather, population distribution by race (Black, Hispanic, Asian ratio), income, street and building structures, position of bars and stations, etc. The highlighted feature of CRIME NABI is temporal and spatial algorithms to model specific patterns, which crime inherently distributes, by hybridization of machine learning, physics, and spatial statistics.
For instance, near-repeat victimization is a well-known cascading trend in which a past crime triggers future crimes in its temporal and spatial vicinity. Their algorithm captures this near-repeat victimization by employing a Green's function scheme on the basis of the self-exciting point process model, and showed good predication accuracy of future crimes (Kajita & Kajita, in press, “International Journal of Forecasting,” para. 1-21). Another merit of this system is that it offers interpretable quantities, such as cascading property and spatially-varying correlation coefficients of crime, which enable us to analyze and expect the causality of phenomena. The Crime Nabi also accompanies a routing engine given the prediction results. The smart-designed route can facilitate to define effective operations for patrol, human resource logistics, and advanced crime analysis.
Mobile App "Patrol Community"
The overview of the AI routines in "Patrol Community" is shown in below. In real time information gathering, data relevance to crime incidents are collected from crime data, weather data, demographic data. Reports by citizens are also saved in the database. These data is fed to "CRIME NABI". The highlighted feature is spatio-temporal analysis to model characteristic patterns which crime events inherently have. For instance, near-repeat victimization is a well-known cascading trend in which a past crime triggers future crimes in its temporal and spatial vicinity. The algorithm captures this near-repeat victimization by employing a Green's function scheme on the basis of the self-exciting point process model, and showed good predication accuracy of future crimes (Kajita & Kajita, in press, “International Journal of Forecasting,” para. 1-21). The output of prediction is sent to the routing engine to propose smart-designed patrol routes. These machine-learning components can assist a crime investigator by providing effective and efficient operations for patrol and logistics of human resource, which is useful for not only police departments but also for citizens engaged with governments.
- 2019/Sep~ Tokyo
We started PoC for iOS Mobile App"Patrol Community" in Tokyo from September 2019. This PoC is supported by the commissioned research by National Institute of Information and Communications Technology (NICT) , JAPAN.
Other several prefectures in Japan
[Selected international infos]
National Institute of Information and Communications Technology (NICT, Japan) from 2018/12-
- Japan Times; Japan considers crime prediction system using big data and AI 2018/06/24
- IoT Lab Selection finalist 2019/02/27
- VLED Disruptive Innovation Award (Oracle Japan Award) 2016
- Linked Open Data Contest (LOD) Data Science Award 2016
- Challenge Open Governance finalist 2016
"Crime prediction by data-driven Green’s function method", Mami Kajita and Seiji Kajita, 2019, International Journal of Forecasting, doi :https://doi.org/10.1016/j.ijforecast.2019.06.005
- International Conference on Spatial Analysis and Modeling 2018/09/08, Tokyo, Japan
- The American Society of Criminology (ASC2018) 2018/11/15, Atlanta, USA
- (to be appeared) Code For Japan Summit 2019
- (to be appeared) ASC2019; The American Society of Criminology 2019/11/16 San Francisco, USA, Crime Prediction by Multivariate Data-Driven Green’s Function Method” in Session Area X. Methodology / 69. Advances in Quantitative Methods
- (to be appeared) ASC2019; The American Society of Criminology 2019/11/13 San Francisco, USA, "Community Policing: The Need of Transition to Cyber Community" in Session "Cyberpolicing and Crime Prevention" collaborations with Dr. J. Lee and Dr. K. Choi
- CEATEC Panel Discussion 2018/10/19
- Cabinet Office 2018/03/29
- Tokyo Metropolitan Police Department Expert Study Group 2018/02/08
Mami Kajita is a Founder and CEO at Singular Perturbations Inc.(SP), and she leads projects of data analysis and software developments for social systems. She earned PhD in statistical physics in 2010, and her profession is statistical physics, especially theoretical research on glass transition and dynamical systems. She is now also a visiting researcher at Center for Spatial Information Science, The University of Tokyo, and studying computational criminology. Mami is also a developer for iOS and software tools.
Resona Kudan Building 5F KS Floor
1-5-6 Kudanminami Chiyoda-ku
Resona Bank on the first floor
1 min walk from Kudanshita station - Exit 6 (Subway Tozai Line / Hanzomon Line / Toei-Shinjuku Line)
The Kudanshita station is the 2nd station after Otemachi station (Tozai line from Tokyo station).