Technology

This system, named the Prophet of Crime,
predicts “when and where future crimes will occur.”

Based on past crime patterns and surrounding conditions, we can predict locations where future crimes are likely to occur.
Furthermore, by suggesting patrol routes that focus security efforts on these high-risk areas based on crime predictions,
we enhance the crime prevention effectiveness of security measures.

Compared to the actual crime occurrence map (left), conventional methods (center) can only provide vague predictions, whereas using Crime Nabi (right) enables highly accurate, high-resolution predictions.

Feature 1: High-Precision, High-Resolution Predictions

We predict crime using various data such as past crime records, demographic information, building attributes, and satellite imagery.

01. Foundational Technologies Supporting High-Precision, High-Resolution Predictions

World's fastest speed with proprietary data compression technology

Our unique data compression technology dramatically reduces calculation costs. While conventional methods require longer calculation times as the amount of input data increases, CRIME NABI achieves dramatically faster speeds. Because we utilize a method that can calculate the cost function optimization process in O(N^0), edge computing is also possible. This technology is also leading to a future where calculations can be performed on mobile devices, drones, and robots to provide optimal security.

02. Technology Supporting the Provision of Crime Prediction Services to the Private Sector

Prediction from Fractional Data Using Transfer Learning

By training a model in an area with crime data and combining it with our transfer learning technology, we can predict crime even in areas where police crime data is not available.
Highly accurate crime predictions are possible even in areas where crime data is unavailable, such as private property owned by private companies.

Feature 2:
Route Planning, Security Operations Planning, and Quantitative Evaluation of Crime Prevention Effectiveness

Based on predictions, we develop security plans utilizing vehicles and cameras. We provide optimal deployment of vehicle and personnel security resources combined with routing algorithms, as well as optimal placement of surveillance cameras.

Furthermore, based on actual security activity data, we can evaluate the crime prevention effectiveness of security measures.

Feature 3: Improved Security Effectiveness

Patrol routes that focus on areas predicted to have a high probability of crime significantly enhance security effectiveness.

Verification in Tokyo, Nagoya City, and Adachi Ward demonstrated results achieving over 1.5 times the effectiveness compared to conventional crime prediction methods. This confirms that the spatial distribution of CRIME NABI's prediction results is more effective than traditional approaches.

Deploy limited resources precisely where they are needed

In fact, implementing CRIME NABI reduced crime by 68.5% in Brazil.

Specifically, in Belo Horizonte, the capital of Minas Gerais—Brazil's second-largest state—the city police organization conducted patrols using CRIME NABI targeting cable theft. We collaboratively verified with local police, universities, and the city government that significant differences emerged between areas where our system was deployed and those where it was not. This finding has also been reported in the form of an academic paper.

Main Research Findings

Research Projects on Crime Prediction: Selected Projects, Research Papers, Patents, etc.

A research team specializing in computational criminology, spatial statistics, computational science, and criminology is developing various crime prediction algorithms.

Selected Projects

NEDO SBIR Promotion Program 2021
NICT-commissioned research "Demonstration-based research and development for solving regional issues through data collaboration and utilization" (December 2018 to March 2021, Final evaluation: highest grade "S")
Tokyo Metropolitan Small and Medium Enterprise Support Center Next Generation Innovation Creation Project 2020 Grant Program
Tokyo Metropolitan Small and Medium Enterprise Support Center, 2020 Second Innovative Service Commercialization Support Project
Ministry of Economy, Trade and Industry Global South Future-Oriented Co-Creation Project Subsidy (2025)

Research Papers

Quantifying Crime Deterrence Effect of Patrol Optimization through GPS Data
Mami Kajita, Daisuke Murakami, Seiji Kajita, Georgia Ribeiro, Genilson Zeferino, and Claudio Beato
preprint (2024)
Fast Spatio‐Temporally Varying Coefficient Modeling With Reluctant Interaction Selection
Daisuke Murakami, Shinichiro Shirota, Seiji Kajita and Mami Kajita
Geographical Analysis (2025)
Spatial process-based transfer learning for prediction problems
Daisuke Murakami, Mami Kajita, and Seiji Kajita
Journal of Geographical Systems (2025)
Compositionally-warped additive mixed modeling for a wide variety of non-Gaussian spatial data

Daisuke Murakami, Mami Kajita, Seiji Kajita and Tomoko Matsui
Spatial Statistics (2021)
Scalable Model Selection for Spatial Additive Mixed Modeling: Application to Crime Analysis

Daisuke Murakami, Mami Kajita, and Seiji Kajita
ISPRS International Journal of Geo-Information (IJGI). (2020)
Crime prediction by data-driven Green’s function method
Mami Kajita and Seiji Kajita
International Journal of Forecasting (2019)

Patents

Patent2020-061937
Patent2019-518001, US, EU, CH, PCT
Patent2018－524169