Risk Management Solutions

As part of our development road-map, we are blazing a new path forward by incorporating machine learning and artificial intelligence (AI) methodologies.  What if you could automate your profiling of high risk shipments, people, and conveyances, and bring those "learnings" to your front end decision support system (e.g. our Operational Risk Management System) in real time?   We are going to merge the strategic back end historical trend analysis and predictive analytics with tactical and operational front end targeting and selectivity systems. Do you desire to be the first customs organization to do this effectively and demonstratively?  We're looking for early adopters.

Our legacy is in building mission‐critical, enterprise class data decision support applications that manages and ingests transaction data from cargo reports (carriers) and declarations (importers).

We build highly configurable expert systems for our customers that drives the risk assessment for a variety of threats (e.g. Revenue Evasion, Narcotics, Terrorism/Security) using a unique library of risk indicators based on both World Customs Organization (WCO) recommendations / leading practices.  Our approach brings a unique risk analysis framework to identify either anomalous or usual and customary commercial practices in order to assess the risks presented by commercial cargo shipments.

We also include a measurable workflow, and collaboration tools for both analysts and inspection teams as an integrated part of the Risk Management process. The result is a powerful and flexible system which captures and reinforces experience and expertise while focusing resources where and when they are needed most.

We are one of the few solutions that has managed to bring "back-end" historical trend analysis and forecasting to front end mission critical decision support with the capability to target and select high risk shipments, people, or conveyance threats.

The diagram above illustrates our approach to machine learning and automated profiling.

Machine Learning and Artificial Intelligence

Using the principles of machine learning and artificial intelligence, we are developing the next generation of transactional analysis, by automating a cyclical framework using the principles of reasoning, knowledge, planning, learning, and perception.

Reasoning

The basis of our approach begins with a business rules management system (BRMS) designed to support decisions for customer missions. Highlights include the following:

  •     Our BRMS runs on a database of transactions or events.
  •     The BRMS manages and vets lookouts, watch-lists, and sanctions.
  •     The BRMS uses a default set of mission centric rules that identify anomalous behaviors in the data.
  •     This process is designed to initiate fast intuitive judgments for making initial logical deductions.

Knowledge

Our Knowledge process determines “What exists” in the transactional data including objects, properties, categories, and relations between these objects, situations, events, and time.

Based on the available transactional data, an ontology is created and a paradigm for decision making is established.

A business object model is established and configured in the BRMS based on the customer environment, domain visibility, and mission.

This is a process that essentially maps the business rules managed in the BRMS to the transactional data.

Example business object models could include: sales and purchase transactions, cargo reports, importer declarations, bills of lading, SWIFT transactions, Automated Information System (AIS) Posits, Advance Passenger Information, and Passenger Name Records.


Planning

Our planning stage initiates 3 core functions.

  1. Risk Scoring:  The rule logic identifies anomalies in the data events or transactions and appends a score to flag and then ranks the events or transactions on a scale of severity and interest.
  2. Case and Workflow:  A configurable decision threshold can then be set to trigger an intervention request to cue investigators or other enforcement teams into action.
  3. Inspection/Investigation Results:  Outcomes of the investigation are collected on a handheld mobile device using pick lists, and standardized reporting templates.

Learning

The outcome of the investigation field report is used as the data selections are made/input into the mobile hand-held device.  This data is uploaded dynamically as it is created.

The application initially determines whether the investigation was successful or not.  If successful, the system automatically tunes and configures the rules that had identified the original anomalies in the data event, - by increasing the interest and severity (I&S) score associated with the rule.

If not successful, the agent reduces the I&S score down.  Over a pre-determined period of time, the rule could turn itself off without any successes to confirm its reasoning.


Perception

Machine perception and awareness is realized through the ability to use input from historical data to deduce aspects and create awareness.

Once initially configured, our technology performs an automated historical trend analysis on the transactional or event data, and will summarize the findings in a strategic intelligence dashboard as suggested profiles.

Any suggested profile can be clicked on by a user to view properties.  A wizard decision flow prompts a user to confirm or negate the profile.

Confirmations become new rules in the BRMS. (User Over-rides do not.)

The Perception Agent learns from user decisions.  Over time, the wizard can be bypassed to allow the perception agent to make automated decisions to confirm profiles based on probability and other factors.


Want to learn more?  Contact us.