As IT investments focus more and more on cost reductions, enterprises are seeking more advanced and strategic use of computer systems through data analytics and optimization technologies based on sophisticated mathematical models. IBM Research - Tokyo has world-class researchers in applied mathematics, especially in the field of data analytics, machine learning, and optimization. Through collaboration with clients, these technologies have had a great impact on real-world problems, such as optimization in steel mill plants.
Competency fields
Machine Learning and Data Mining
Our research focus is on methodologies and frameworks for deriving insights into businesses and services from the huge volumes of data now available from maturing IT infrastructures, and linking these insights to actions. We are studying fundamental analysis methods such as anomaly detection and risk-sensitive data analytics, and also obtaining many results by applying these methods to time series data in manu-facturing and CRM data, leveraging the merits of our proximity to advanced companies and markets in Japan.
Optimization Technologies
Standard optimization techniques do not easily apply to complex problems, having uncertainties, that companies and societies faces today. We solve these difficult problems by constructing new models, by applying existing models in new ways, by designing new algorithms, and by developing other new approaches.
Optimized Production Scheduling
We are tackling various discrete optimization problems related to production design and operations scheduling. Thanks to the development of optimization techniques and to increases in computing power, computers can now handle some very large problems that even human experts could not handle, and in other cases, computers can sometimes outperform human experts with higher quality solutions. We, jointly with the Watson Research Center, have been developing a library of optimization solutions called PDOS (Production Design and Operations Scheduling) for steel makers. PDOS is composed of six optimization modules: Inventory Application, Slab Design, Plate Design, Melting Shop Scheduling, Hot Mill Scheduling, and Finishing Line Scheduling. The tools have already been delivered to several leading steel makers in Asia.
Optimization in the Steel Industry
Text Mining
We are developing text-mining technologies, such as systems to analyze the "voices of customers" as found in contact center logs and webpages, and using these results to seek insights that can guide corporate activities. Combining natural language processing technologies, fast indexing technologies, and interactive mining views, we developed a text mining tool called IBM TAKMI, which has now been productized as the IBM Content Analytics.
Social Analytics
Various communication tools such as social network systems (SNSs) and blogs are actively used even in companies. To analyze these tools, a new methodology considering both the content (text) and network structure (person-to-person relationships) is needed. We are working on developing a suitable analytic framework and on creating advanced applications (such as content recommendation systems and advertisements).