As the “Amazon of the Defense Department,” the Defense Logistics Agency manages multiple supply networks, over five million line items and provides billions of dollars in goods to the military services and other partners, including the Federal Emergency Management Agency and NATO. Logistics Information Services catalogs each item using the Federal Logistics Information System for Defense Department personnel, contractors and others who do business with the U.S. government.
In 2020, Logistics Information Services partnered with DLA Research & Development’s Defense Logistics Information Research program led by Dr. Senthil Arul on the FLIS Data Cleansing Project, ushering the system into the data-driven digital age of automation and machine learning.
Developed in the 1960s, FLIS stores essential supply item details including national stock numbers, item names, manufacturer and supplier information, parts numbers and performance characteristics (e.g., material, dimensions, surface treatments and color). Logistics Information Services recognized that it was time to transform the antiquated system which became increasingly inefficient over the years. System errors and inconsistencies led to item duplication, increased costs and impeded readiness.
The project began with a thorough data cleansing, ensuring errors and inconsistencies didn’t enter the upgraded FLIS environment. A new modernized algorithmic base uses machine learning and other analytical techniques to assess multiple data sets and identify the right data – eliminating past challenges.
“Improving data quality and structure – two key components of enterprise data management – is critical to the maintainability and usability of the catalog,” said Kristen Cheman, LMI senior vice president, digital and analytic solutions.
As the project enters Phase III, the team is implementing and executing data cleansing actions including the application of natural language processing techniques to translate free-text characteristics data into structured coded replies and ensuring consistent structure across item names.
“By harnessing the capabilities of Natural Language Processing, we gain the ability to unlock the true potential of our data—transforming unstructured information into structured insights ensuring accurate and actionable results,” Dr. Arul explained.
FLIS Process Lead Stacey Navarro said the partnership with DLA R&D and LMI has been invaluable to the DLA FLIS Characteristics team.
“Their advanced tools and technologies to identify data issues allowed us to see the magnitude of those issues in new ways. What is even more impressive is their ability to gain a deep understanding of our complex data and processes, ensuring tools recommend quality-based cleansing actions,” Navarro said.
“This has not only been a breath of fresh air but also increased our confidence in FLIS results. It allowed a rare opportunity to brainstorm from a fresh perspective, further providing better ways of ensuring the system’s longevity.”
DLA R&D leads several projects dedicated to modernizing and strengthening the agency’s legacy systems. Learn more about R&D’s innovative programs.