|
|
 |
 |
|
| |
|
|
|
| |
|
| |
Recent Data Mining Advances of Aviation Safety Data
September 22, 2006 / 11 a.m. ET
Presented by: Dr. Irving C. Statler and Dr. Ashok N. Srivastava
This seminar describes advances that have been made in extracting
information efficiently and reliably from large, distributed, multiple,
heterogeneous sources of aviation safety data.
The Information Sharing Initiative (ISI) was started in June 2004
building upon work conducted under the Aviation Safety and Security
Program. The ISI evolved in response to the industry's need for the
advanced technology and infrastructure to enable a distributed archive
of Flight Operational Quality Assurance (FOQA) and Aviation Safety
Action Program (ASAP) data across U.S. air carriers. Those archives have
now been established and demonstrated to be operational. The challenge
is to develop analytical tools to exploit these sources of information
on events or trends that could compromise the safety of the air
transportation system.
Image to right: Safety data from across nine airlines is currently
integrated and can be analyzed by participating operators using new
capabilities developed under the Information Sharing Initiative. The
sources of data and the number of participating Airlines is expected to
expand as new data mining capabilities are further developed and
implemented. Image credit: NASA
The presentation includes a technical discussion of recent algorithmic
advances made to address key challenges in text mining and data mining
of heterogeneous data sources. Several new areas of algorithmic
developments are discussed, including the use of Support Vector Machines
for automatic text classification, the Inductive Monitoring System and
Orca for automatic detection of anomalies in continuous data streams,
and "sequenceMiner," a new algorithm based on ideas from bioinformatics
that discovers anomalies in high-dimensional symbol sequences. The value
of these advances is shown through the use of aviation data as well as
data from Space Shuttle systems. The opportunities posed by the
aerospace (and potentially broader) community for these new advances
include capabilities to:
- Automatically acquire, integrate, and analyze very large amounts of
data from disparate sources (continuous digital, discrete digital,
analog, and textual) in order to detect systemic trends or anomalies in
a timely manner for corrective action.
- Conduct targeted assessments across large and complex operational
systems to measure the impact and effectiveness of new technologies
and/or operational changes introduced into the system.
- Provide a flexible system architecture that is scalable and
adaptable to meet requirements of a broad range of operational users.
+ See Full Technical Seminar Series Schedule
|
|
|
