Data Mining

Key Member

Teruhisa Miura

Project Objective

The huge amount of data is being produced from various areas. In order to extract useful knowledge from large data, there exist theoretical problems to be solved. We focus on solving these problems under limited resources such as time and memory. Our approach is to apply search methods to extract useful knowledge from huge database. We aims to propose a new search method, which is appropriate to search in huge state-space. The target domains are genome sequence data.

Current Status

Research results obtained so far are as follows:

Future Direction

This project aims to propose a new search method for huge database.

So far, there has been no study about search methods, which can be applicable to huge state-space. Conventional search is for search problems that its state-space can be stored on main memory and its next state can be easily generated. The Genome sequence walking can be regarded as a search problem whose state-space is composed of huge data on database. For generate of next state, we have to compare the query fragment with every fragment in database to find which fragment overlaps with the query fragment. This makes the cost of generation of next states very high. In this situation, we need new measure to evaluate effectiveness of search methods. Study on search methods for database become important, because needs to extract useful information from huge data is increasing more and more.

References

1.
Yasuhiko Kitamura, Hideyuki Nakanishi, Tetsuya Nozaki, Teruhisa Miura, and Toru Ishida, ``MetaViewer and MetaCommander: Applying WWW Tools to Genome Informatics,'' Genome Informatics 1996, Universal Academy Press, pp. 137 - 146, 1996.

2.
Teruhisa Miura and Toru Ishida, ``Stochastic Node Caching for Memory-bounded Search,'' National Conference on Artificial Intelligence (AAAI-98), pp. 450-456, 1998.


Top
Page
Free
Software
Publications Members Location Related
Servers