NEC develops Artificial Intelligence technology that makes decisions and plans based on predictions using Big Data
Tokyo, November 2, 2015 – NEC Corporation today announced the development of a “Predictive Robust Optimization Framework Technology,” enabling software to make decisions and plans based on predictions, as part of its artificial intelligence (AI) technology for advancing Big Data analysis.
With the popularization of the Internet of Things (IoT), demand is rapidly growing for analysis and predictions that utilize Big Data gathered from everyday environments. In 2012, NEC developed a “heterogeneous mixture learning technology,” that achieves the automation of highly accurate, large-scale predictions; such as predictions of supply and demand for energy, water and food resources. This has helped to increase the efficient use of resources, to predict the supply and demand for warehouse stocks, thereby helping to streamline logistics management, and to predict the demand for various products, thereby helping to increase the sophistication of retail store management.
“NEC’s Predictive Robust Optimization Framework Technology makes advanced decisions, such as the formation of strategies and plans that would conventionally be carried out by humans, through software that utilizes predictions created by NEC’s heterogeneous mixture learning technology,” said Akio Yamada, General Manager, Knowledge Discovery Research Laboratories, NEC Corporation.
When applied to water distribution planning based on water demand predictions, the new technology was successful in creating a highly accurate distribution plan, achieving a 20% reduction in the amount of power consumption required for purifying and distributing water.
In addition, when the technology was applied to product pricing based on product demand predictions, it was successful in producing a price strategy that would improve store sales turnover by 11%; a strategy that the technology produced both automatically and within the space of less than one second.
NEC plans to achieve commercial use of this technology during fiscal 2015.
Key features of this new technology are as follows:
- The new technology produces plans that are both highly efficient and low-risk with respect to prediction errors.
- By analyzing patterns of typical prediction errors using NEC’s original algorithms, and fusing the results of this analyses with mathematical optimization technology, NEC’s Predictive Robust Optimization Framework performs optimizations in view of how the predictions may be incorrect. In this way, the framework calculates and makes plans that consistently produce highly effective results, and for which the risk of loss or damage arising due to prediction errors is low.
- The Predictive Robust Optimization Framework technology produces optimal plans at ultra-high speed, considering the interrelationships of a large quantity of predictive formulas.
- By the use of NEC’s original combination optimization algorithms, the technology is able to efficiently search through a large volume of combinations while at the same time considering the relationships of numerous predictive formulae, and calculate optimal strategies and plans at ultra-high speed.