Skip to content
This is NOT an official site of the City of New York. Click here for official AI reports.

NYC Office of Technology and Innovation

NYC - DTPR AI Register

AI-Assisted Crime Pattern Detection for NYPD Analysts

What it is for

Safety & Security
  • Patternizr supports crime analysis by detecting potential crime patterns across burglaries, robberies, and grand larcenies, enabling detectives to consolidate investigative efforts around linked incidents.
Enforcement
  • The tool assists NYPD enforcement by helping analysts propose crime patterns for a pattern identification module, which may then direct detective resources toward linked crimes.
Run by
New York Police Department (NYPD)
Where
No fixed location
Kept
Not stated by the NYC Office of Technology and Innovation.
Shared with
Accountable organization
Your copy
You cannot see the data it holds about you. What you can do

What it is for

Safety & SecurityEnforcement

Patternizr is a machine-learning tool used by New York Police Department analysts to identify groups of similar crimes — such as burglaries, robberies, and grand larcenies — that may be connected to a single offender or pattern. It compares features of each crime report (location, time, method, suspect details) and returns a probability score indicating how likely the complaint belongs to an existing pattern. A human analyst must review all results before any investigative action is taken; the system does not make autonomous decisions.

What it collects and what happens to it

Data taken in

Sensitive personal information
Identifiable data
  • Complaint records include suspect information as one of five feature types used to compute crime-to-crime similarity. Training data also includes a portion of complaint records where the same individual was arrested for multiple crimes of the same type.
About a place
Anonymized data
  • Each complaint record includes the location of the crime, used as one of the five similarity features (location similarity). Geographic space is listed as a population type in the register.
About behaviour
Anonymized data
  • Crime complaint records include the date and time of occurrence, crime subcategory, and modus operandi — behavioural and categorical features of the criminal act used to compute categorical and unstructured-text similarity scores.
Operational data
Anonymized data
  • Approximately 10,000 manually identified crime patterns per crime type (burglaries, robberies, grand larcenies) from 2006–2015 were used as training examples. These historical pattern records are administrative/operational data used to train the similarity models.

Processing

Classification & Prediction
  • Patternizr uses separate machine-learning models trained for each of three crime types (burglaries, robberies, grand larcenies). Each model computes five crime-to-crime similarity features and returns a probability score that a candidate complaint is part of an existing crime pattern.

What it does

Deciding (Analytical AI)
Human decides
  • Patternizr scores crimes by computing five similarity features (location, date-time, categorical, suspect, unstructured text) and returns a probability that a complaint is connected to a pattern. Human analysts evaluate the output and decide whether to propose a formal crime pattern.

Outputs

A recommendation or prediction
Anonymized data
  • The system outputs a probability score indicating how likely a complaint is connected to an existing crime pattern. It also filters non-normal trends into a spreadsheet and displays year-over-year crime counts with non-normal trends. These are advisory outputs reviewed by human analysts.

Run by

New York Police Department (NYPD)
  • The NYPD deploys and operates Patternizr to support crime analysis and pattern detection. Analysts and detectives use the tool's outputs to guide investigative consolidation.

NYC Algorithmic Tools Compliance Report — Patternizr

Built by

Not stated by the NYC Office of Technology and Innovation.

Kept for

Not stated by the NYC Office of Technology and Innovation.

Shared with

Not available to me
  • The probability scores and candidate crime lists produced by Patternizr are internal NYPD analytical outputs. They are not available for individual members of the public to access.
Available to the accountable organization
  • NYPD analysts and detectives have access to the outputs of Patternizr for the purpose of crime pattern analysis and investigative consolidation.

Stored

Not stated by the NYC Office of Technology and Innovation.

How to read the colours

Can it identify you?

Anonymized data
Data about people with the link to who is broken. Stripped of identifiers, blurred, aggregated, or noised so this system can’t reasonably tie a record back to an individual.
Pseudonymous data
Each person’s data is tied to a token (hash, ID, template) that lets this system recognise the same person across events, but the token itself doesn’t reveal a name. Reidentification is possible with extra information.
Identifiable data
The data either contains a direct identifier (name, address, account name, recognisable face or voice, plate number) or carries a token this system uses to look up legal identity during processing.

Who completes the loop?

Human decides
This mode suggests; a person decides what to do next. The AI is always advisory — a human is in the loop on every decision. Example: a triage tool ranks cases for a clinician who chooses which to see first.
Human executes
This mode decides; a person carries out the result. Example: an optimizer plans the day’s trash-collection routes, and drivers run them.
Autonomous
This mode decides and acts on its own. No person reviews each decision or carries out the resulting action.

Definitions from the DTPR standard. Amber is about your data, violet about who decides. The fuller the shape and the deeper the colour, the more identifying the data or the less a person is involved.

What you can do

Ask about this system

Questions go to the NYC Office of Technology and Innovation, not the vendor.

Your rights

  • Right to Algorithmic TransparencyThe existence and general operation of Patternizr is publicly disclosed through the NYC Algorithmic Tools Compliance Report. Members of the public may review the register entry at the source URL. The system compares crime features across five similarity dimensions and returns a probability score reviewed by a human analyst.
  • Right to a Human ReviewAll outputs from Patternizr are reviewed by a human analyst before any investigative action is taken. The system does not make autonomous decisions; analysts must evaluate whether candidate crimes are genuinely connected to a pattern before patterns are formally proposed.

Risks and safeguards

  • Civil liberties harmCrime-pattern detection in a law-enforcement context may contribute to over-policing of certain geographic areas or populations if pattern boundaries are drawn inequitably.Safeguard: The system is explicitly informative only — a human analyst must evaluate all candidate crimes and propose patterns to a separate pattern identification module before any investigative action is taken. No automated enforcement decision flows from the model output.
  • Reputational harmIndividuals whose complaints are flagged as part of a crime pattern may be subject to intensified investigative attention based on probabilistic similarity, not confirmed guilt.Safeguard: The tool outputs a probability score, not a determination of guilt or an arrest recommendation. The register states the tool requires a human user to evaluate the output data to see if complaints identified as similar are, in fact, connected to a pattern.