CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
Online Citizen Access Portal for Alabama Food Assistance
Program
Case management system for the
Alabama Office of Radiation Control
Custom business rules engine and workflow builder and manager
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
96% accuracy rate in benefit payments. Considerably higher than most major benefit programs.
96% 4%
20% payment errors are actually due to underpayment .
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
OVERPAYMENT
$117 MILLION / MONTH
IN OVERPAYMENT OF
BENEFITS IN THE US
MEALS
ENOUGH TO FEED
135,252 FAMILIES OF 4
FOR AN ENTIRE MONTH
SAVINGS
POTENTIAL SAVINGS
OF $1.4 BILLION / YEAR
NATIONWIDE
MEALS
ENOUGH TO PROVIDE
592 MILLION MEALS
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
01.
Don’t rely on a single source of data
This system pulls data from multiple sources such as the US Census and state
FNS records.
Ensure data is easy to add
New data can be added to the system through a powerful user interface. This creates a system that prevents stagnation through frequent updates.
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
MANY MORE
US CENSUS
ALDHR FNS
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
02.
Value the experience of your users
The system should display data in a way that allows users to “follow their hunches”. The system should support, not replace the user.
Capture user feedback to improve prediction
As users interact with the system, capture their input and use it to improve analysis over time.
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
03.
The world changes at a rapid pace
The system should provide tools to implement changes in an efficient manner. The more high-level the interface the better, you should limit the involvement of “super hero programmers” when possible.
Embrace the diversity of your users
People interact with the world in many ways. Don’t force a user to use your system through a single view. Instead, allow many different means of exploring the system.
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
04.
Don’t assume users are data scientists
Guide users to making good decisions on how they visualize and interact with the data. Just because they can do it, doesn’t mean they should.
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
05.
Use concepts familiar to users
Use interface design elements, terminology and functionality that is familiar to the end users. Avoid complex jargon when designing the system.
Spend significant effort on user experience
The most accurate predictive system in the world is useless unless people use it. Special effort should be spent on improving the “everyday” user’s experience.
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
06.
Fraud implies negative intent
While it is imperative that data is monitored for invalid and fraudulent activity, looking for what is different versus what is bad allows the system to identify issues that can provide value in improving society.
Identify and communicate subtle differences
As transactions and data change based on a variety of factors, ensure system users are alerted in a timely manner.
Food Assistance Applications in Mobile County, 2013
600
500
400
300
ANOMALY CAUSED BY OIL SPILL
200
100
0
White
Black
Asian
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
Keep the
“human in the loop”
Combine analysis with prevention
Look for anomalies, not fraud”
Make it flexible, approachable and sustainable
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU
@UACAPS
UACAPS caps.ua.edu
205-348-6835 csims@cs.ua.edu
CAPS | PREDICTIVE ANALYTICS FOR PROGRAM INTEGRITY | CAPS.UA.EDU