Kohakusha


harness the power of your data



ABOUT

Delivering cutting-edge solutions to domestic and overseas clients' data-mining and analytics challenges, Kohakusha is strategically positioned to take a leading role in Japan 's rapidly expanding market for business intelligence.

COMPANY

You have at your fingertips more data than ever before. You can access unprecedented levels of information about your markets, customers, business practices and technology. But how can you put this data to work for you?

Kohakusha can turn your data into knowledge to give you competitive advantage and bottom line results. We specialize in providing consultancy and bespoke development based on cutting edge statistical analysis, data mining and predictive modeling technology.

Demand for data mining and modeling is booming as volumes of information in data warehouses, data marts, and on company servers increase at exponential rates. It is no secret that companies which can make sense of this data are able to make more intelligent, more profitable business decisions.

Kohakusha offers a comprehensive range of statistical analysis, development and consulting services to provide you with unmatched knowledge access solutions. We can help you make sense of your data. And it doesn't end there. We can use our expertise to build accurate predictive models of your business environment, future market conditions, customer behavior, and more.

Kohakusha is unique in its consultative approach to data mining and modeling challenges. Business intelligence software providers tend to develop their product lines, and then shoehorn customers' problems into incompatible technology – quick-fix solutions that have often come at the cost of predictive accuracy.

We believe that data miners must first understand their clients' individual goals and the specific characteristics of the phenomena to be modeled. Only with this understanding can the most appropriate conceptual frameworks and techniques be selected, and the task of developing mining and modeling solutions uniquely suited to their purpose begun. When predictive accuracy is the primary goal, we believe that no other approach can be as effective.

With this conviction firmly in mind, Kohakusha's goal is to provide statistical analysis, data mining and modeling solutions that enable our clients to leverage business intelligence into decisive competitive advantage.

With the competencies and expertise we have developed, with our strong partnerships in the Asia Pacific region, Kohakusha is positioned for success in the data mining and modeling boom which we confidently predict for the 21st century.

SERVICE

Consulting

Our expertise in developing knowledge access solutions allows us to provide a broad range of services, from classical statistical analysis to innovative predictive modeling consultancy. Corporate clients have sought our specialist knowledge in:

- Market and customer analysis
- Futures modeling
- Business Process Modeling
- Information requirements for effective database mining
- Effective implementation of tools for response modeling, pattern recognition, clustering and segmentation
- Diagnostics and interpretation of in-house servers
- Discrete choice survey instrument design

Training

As a complement to our consultancy services, Kohakusha offers statistical modeling and data mining training programs. These include seminars and hands-on practical training courses either at the client's site or at our Tokyo offices.

Programs in Data Mining with Tree-structured Mining Tools, and Predictive Modeling with Automated Regression Tools are general courses designed to equip non-specialist users with the knowledge needed to get the most from these technologies. We can also create bespoke courses tailored to clients' specific requirements.

DEVELOPMENT

Bespoke Development and systems integration

To get the most out of your data you need more than simple software tools. Optimal deployment of data mining and modeling technology requires specialist knowledge, software development and systems integration. Take advantage of Kohakusha's experience to ensure you derive the maximum advantage from your business intelligence software.

We develop products with a clear understanding of our clients' goals, a wealth of expertise in the knowledge access technology and according to the specific characteristics of the data challenge in question.

In developing appropriate mining and modeling solutions, our experience suggests that two types of technology may be especially applicable for many clients' needs.

Tree-structured mining tools

Probably the most popular commercial data mining technique, the great advantage of “decision trees” is the readily understandable graphical representations of the models they generate. Such models have a proven record of success in applications such as customer profiling, targeted marketing, fraud identification and risk management. Well known decision tree algorithms include Classification and Regression Trees (CART), Chi Square Automatic Interaction Detection (CHAID), and QUEST.

A decision tree algorithm is the most appropriate mining technology where transparency and ease of use are priorities. For the same reasons, a central feature of Kohakusha's solutions involving such technologies is intuitive Windows-based interfaces.

Other features that will be included in any decision tree solution will be:

- Reliable stopping rules – an optimal pruning strategy
- Overfitting checks – self-validation procedures
- Intelligent error handling and handling of missing values

Automated predictive modeling

Taking advantage of advances in automated regression modeling dramatically shortens the once trial-and-error process of building accurate predictive models, saving time and money.

Best-fit variable transformations can be mechanically determined with extreme precision by contemporary technologies. Given a target variable and a set of candidate predictors, they can automate all stages of model development, including:

- Establishing relations between predictors
- Transforming predictor variables with non-linear relationships to the target
- Isolating relevant and irrelevant predictors
- Handling missing data with nested variables
- Comprehensive and continuous self-verification to prevent overfitting

These features combined with easy-to-use interfaces and where possible graphical representations of the modeling process can turn what was once a painstaking, stop-and-start process, into an intuitive, fluid progression.

CONTACT

5F Risona Kudan Bldg
1-5-6 Kudanminami
Chiyoda Ku

conformalstring _at_ gmail.com