Value Creators
Sawako Mikami

NEC has developed the world's first "heterogeneous mixture learning technology" and is creating many different types of value through the highly accurate predictions this technology enables.

What do you focus on as a data scientist whose mission is to create new value by utilizing big data?

Sawako Mikami
Sawako Mikami
Assitant Manager
Big Data Strategy Division,
Knowledge Discovery Research Laboratories

Creating new value by utilizing big data is a trending topic right now. One of the areas gaining attention prediction and forecasting, a field in which we have been unable to achieve a satisfactory level of accuracy up until now. NEC was the first in the world to develop "heterogeneous mixture learning technology," which is a technology that we can use to carry out highly accurate forecasting by using big data. This technology is already achieving results in a range of applications such as power demand projections for the energy industry and product demand projections for the distribution and retail industries.
I am currently working on product demand projections. As a data scientist, my most important job in helping customers utilize big data is to clarify exactly what they will use the data for. The data that must be collected differs depending on the goals of each customer and the issues they face. In order for a data scientist to accurately determine what kind of data to use, what to analyze and what to predict, the scientist must communicate with the customer and understand his or her situation.

How has forecasting accuracy and the work of data scientists changed with the development of the NEC's proprietary analytics engine?

The biggest feature of the heterogeneous mixture learning technology is that it can automatically discover patterns and hidden rules in the vast amounts of data that it processes. For instance, in the product demand projections that I'm currently working on, demand depends on many different conditions, such as weather, day of the week, time, temperature, whether an event is being held and so on. When these conditions are combined with the hundreds and thousands of products and stores located around the country, we end up having to carry out a huge amount of data partitioning in order to create an accurate forecasting formula.
For the data scientists faced with the job of partitioning all this data, an analytics engine that automatically creates the necessary forecasting formulas is revolutionary. In addition, because the conditions upon which the data partitioning is performed and the contents of the forecasting formulas are easy to understand and visible, the engine can also be used to improve operations post deployment. Heterogeneous mixture learning technology has helped to both increase the accuracy of forecasts and the speed of analysis.

What is the future of prediction and forecasting, and what are your thoughts on the future roles of data scientists?

The evolution in analysis technologies, the increase in sensing devices, and the development of IoT will lead to the mainstreaming of big-data-based prediction and forecasting in fields such as public safety, including urban infrastructure and traffic, facility deterioration and resource demand predictions, and health care. NEC will continue to use advanced big data analytics to create new social value and support business growth. Personally, I will try to accumulate more experience and knowledge, and proactively absorb new technologies and trends to become an even better data scientist. For example, when I am shopping at a convenience store in my free time, I tend to look at the product layouts and the available products. I look at things as both a data analysis professional and as a consumer. I think that both of these points of view are equally valid and useful, and I would like to use them to make the best proposals for our customers.

Reading NEC Vision Book 2017

More information about NEC's vision for social value creation and the steps we are taking to achieve this vision can be found in NEC's Vision Books.