NEC develops AI technology that supports new policy planning based on past data～ Contributes to pre-identification of product defect factors and analysis of customer purchasing behavior ～
Tokyo, February 25, 2022 - NEC Corporation (NEC; TSE: 6701) today announced the development of “Rule Discovery based Inference" an artificial intelligence (AI) technology that presents specific conditions for improvement by deriving the causes of events and the conditions under which they occur based on past case data, and enables support for policy planning in easy-to-understand language.
This technology contributes to the pre-identification of product defect factors and the analysis of customer purchasing behavior in areas such as manufacturing, retail and finance.
Data-based analysis of current conditions, such as product quality control and customer analysis, is conducted in a variety of industries. In recent years, AI has been used to predict the quality status of products and to predict customer purchasing behavior. Specifically, AI presents future forecasts and the factors that will lead to those forecasts, and improvement measures and countermeasures are formulated based on expert experience.
However, as labor shortages worsen and skilled engineers and professionals age, there is a need to pass on technologies, experience, and know-how, as well as to use AI to provide total support, including policy planning and future forecasts.
In order to address these issues, NEC has developed this technology as part of its promotion of AI that can co-work with people and become prevalent throughout society.
Features of the technology
This technology learns past cases as "correct data," finds results, the factors leading to results, and the conditions under which they occur. By establishing rules for "what factors occur under what conditions," it is possible to show the conditions for improving each factor.
For example, in the factor analysis of product defects in the manufacturing industry, many factors, such as the compounding of raw materials and the setting of treatment equipment, affect the rules that are established, and the number of rules can become enormous. Therefore, it is not realistic to comprehensively investigate these rules and derive the improvement conditions for each factor.
This technology explores fewer and more precise rules by prioritizing each rule in its own way. Specifically, it builds a group of rules by learning the data that caused product defects and the data that did not cause defects. From there, each rule is prioritized in its own way, and by applying parallel calculation techniques, it is designed so that sufficient rules can be sorted with a small number of calculations.
In a proof-of-concept using open data, NEC confirmed that nearly 50 rules were required with existing methods to cover an entire case, but this technology could achieve results with approximately a dozen rules. This makes it possible to derive more precise rules with fewer rules than conventional methods.
Case 1: Analysis of manufacturing defect factors in the manufacturing industry
Technology verification in product defect factor analysis not only allows us to identify the factors affecting defective products, but also allows us to find out the possibility that factors not previously assumed by experts may affect defective product generation, such as "failure with a probability of 80% when the temperature of the material is higher than 100 degrees Celsius and the pressure of the equipment is higher than 20hPa."
Case 2: Customer Purchasing Behavior Analysis in Retail
Technical verification in customer purchasing behavior analysis confirmed that based on historical customer purchasing data and store policy data, this technology was able to visualize what triggers change from new customers to prime customers and the terms of such changes, and support for the formulation of specific measures.
For example, "when the number of purchases of product A is greater than 10 and the number of visits to the store is greater than 50 times, a 90% chance of purchasing product B can be presented," and specific factors that lead to results and the conditions under which they occur can be presented, making it easier to formulate measures.
Going forward, NEC will continue to support human decisions in a complex and rapidly changing society through the development of AI that people can trust and accept.
A clear understanding of AI thought pathways: Rule Discovery based Inference
About NEC Corporation
NEC Corporation has established itself as a leader in the integration of IT and network technologies while promoting the brand statement of “Orchestrating a brighter world.” NEC enables businesses and communities to adapt to rapid changes taking place in both society and the market as it provides for the social values of safety, security, fairness and efficiency to promote a more sustainable world where everyone has the chance to reach their full potential. For more information, visit NEC at https://www.nec.com.
NEC is a registered trademark of NEC Corporation. All Rights Reserved. Other product or service marks mentioned herein are the trademarks of their respective owners. © NEC Corporation.