Data mining comprises the technology, systems and processes used to discover knowledge within large-scale data sets. The core members of the Data Mining Laboratory are engaged in development of basic technology, applications, processes and practical testing of data mining for business uses. The DMLab conducts collaborative research with companies in a broad spectrum of industries. The applications developed through this research have been adopted by several companies, with outstanding results.

The following describes research the DMLab is currently pursuing, as well as examples of previous work.

◆MUSASHI (data mining-oriented open source platform)

A group of commands developed for processing large-scale data sets, MUSASHI will be indispensable during future years when mining business-related data. MUSASHI is capable of processing XML table data and plain text data with a table structure. Being open source, MUSASHI can be freely downloaded and customized. When combined with other free or paid software, it becomes an extremely powerful analysis tool.

◆E-BONSAI (category time series analysis system that uses character string analysis technology)

character string analysis technology): The character string analysis system BONSAI was developed for data mining in the field of genome analysis. The system extracts distinctive character substrings existing in specified groups within long character strings. Having hit upon the idea of expressing consumer behavior patterns as character strings, we developed E-BONSAI as a tool that applies character string analysis to analysis of time-series purchase patterns for business application. Employing E-BONSAI, the group managed to construct a model capable of accurately forecasting consumer brand switching behavior. When tested in sales promotion experiments in a retail environment, sales increased and useful findings were obtained.

◆Graph mining applications in business

Graph mining is a system that extracts distinctive subgraphs from graph-structured data expressed as 'links' and 'nodes'. We developed a framework for application to business of the AGM (Apriori-based Graph Mining) developed at the Washio Laboratory in Osaka University, and obtained useful findings in experiments focusing on beer and other products. Building on these findings, we developed novel sales promotion strategies that brought sales increases not only in the target products but also in related products (such as fresh fish and others, when beer was the original target), thereby providing the foundations for development of new sales strategies.

◆C-MUSASHI (CRM system that uses WEB Log Data and Customer Data)

C-MUSASHI is a customer relationship management (CRM) system that runs on MUSASHI and employs data mining technology. Using customer data obtained from real stores or through Internet shopping, C-MUSASHI formulates effective sales strategies from past purchasing trends. As it is aimed at companies managing a number of stores, C-MUSASHI includes a store management system that provides a store manager with findings for development of store strategies appropriate to the varying environments and customer behaviors found in different stores. The system can be used in a variety of industries, including foodstuffs, miscellaneous daily goods, electrical appliances, and automobiles.

◆CODIRO (consumer survey system that uses data mining)

CODIRO is a system that surveys consumer behavior by organically fusing not only in-house customer data, but also mobile sites data, sales promotion databases from Internet survey companies and manufacturing companies, as well as a variety of other data. By measuring consumer recognition of advertising and brands and accurately relating this to purchase behavior, CODIRO makes possible highly accurate consumer surveys. The Data Mining Laboratory has published work describing the relationship between the consumer recognition process and purchase behavior elucidated through collaborative advertising experiments with food companies.

◆Knowledge discovery from persuasive communication

Some businesses spend large sums of money on debt collection. Debt collection relies heavily on the human element, with results hinging on the negotiation skills of the individuals involved. Through analysis of the negotiation process in the debt collection industry, we developed a system to analyze processes for effective persuasion. As the system permits scientific analysis of negotiation in addition to clues provided by intuition and experience, it is a valuable tool with great promise for a broad range of future applications. In a previous paper, we describe our success in discovering useful findings through analysis of debt collection at some communications company.

◆PRISM (optimal pricing system that employs data mining)

PRISM is a system that employs data mining technology to find optimal prices. On the basis of a range of consumer purchase history data, PRISM extracts optimal pricing patterns across a number of categories. PRISM provides a scientific approach and system for effective and profitable price setting in the retail industry, a sphere in which meaningless price reductions are commonplace.

◆Consumer movement analysis using RFID data

RFIDs are extremely small IC tags employed in various fields including logistics, where they are used for positional confirmation. Using data obtained from using RFID to track in-store customer movement, we are attempting to discover useful information contained within the path of individual customers around a store. By intricately analyzing purchase behavior, a consumer purchase process, we are attempting to obtain useful knowledge that will improve the efficiency of sales promotion activities.