Nomura Securities deploys AI from Fujitsu to improve data quality
Nomura Securities will be deploying an analytical machine learning product from Fujitsu in June 2017.
The deployment is expected to provide improvements to data quality in areas where conventional methods of ensuring data quality has reached its limits. The deployment aims to utilize both actual data and AI to improve data quality.
Alongside the processing performed by securities systems, Nomura has needed improvements in the quality of data recorded and stored day by day through such processes as a variety of manual entry tasks and internal business processes. Previous methods ensuring data quality, were limited in their ability to grasp human input errors and the patterns of occurrences when data content and method of input deviated from the norm.
Fujitsu and Nomura Securities carried out a joint trial focusing on the occurrence tendencies and frequencies of past data. Using machine learning core technology and acceleration technology developed by Fujitsu Laboratories Ltd., the two companies verified whether this system can autonomously analyze data simply run through the system, without any prior training, after separating it into ordinary patterns and patterns that deviate from the norm (anomalies).
There had been cases in which manually checking large volumes of operational data in their entirety was impractical, meaning that ordinary system checks were not comprehensive. In this trial a few dozen records that deviated from the norm were separated out from records numbering in the tens or hundreds of millions, including a few records showing patterns that even experts could not have recognized, enabling new discoveries. Because the analytical AI can quickly detect patterns that differ from the everyday norm, this system can significantly improve the efficiency of operations. Moreover, improvements in analysis accuracy can be expected via the building up a store of new discoveries from the detected patterns as expert knowledge, leading to ongoing enhancements to data quality.
Previously, the creation of test cases also had to rely on expert knowledge, but applying analytical AI allowed for the system to autonomously train itself on the operational data that is produced day by day, reflecting new data patterns in test cases. Because it can comprehensively extract data patterns that are highly important to operations, while at the same time omitting unnecessary data patterns, enabling efficient test validation, this system can contribute to improving test quality and productivity.
Given the trial results, Nomura Securities plans to deploy data analysis AI based on these technologies. In addition, noting the general applicability of the deployment at Nomura Securities, Fujitsu plans to offer it in other industries as well.