Abstract
Condition search is mainly used by the data retrieval technique. Since it is difficult to judge which is adjusted when there are two or more search conditions, we have to repeat adjustment re-search repeatedly. Furthermore, depending on a database, even a condition setup may become difficult if the searcher has not familiar with the database. A real estate housing database is considered as an example. The conditions of a housing of wanting to live have a house rent is cheap, a building is new,. a housing space is large, close to a landmark, etc, and the housing which fills these all is the most desirable. However, the concrete value of an ambiguous expression, such as “cheap”, “new”, “large”, “close”, changes greatly with areas which the housing locates. The searcher has to learn these relations through searching. The same problem exists in many fields including job information search and used car search.
In order to solve the above-mentioned problem, new method was proposed that, display feature of information in 2 dimensional map using Self-Organizing Map technique, clustering data on the map by index data, and adjusts conditions visually on the map. As a result of experiments using actual real estate housing data, it was shown that this method is especially effective to look for information on the ambiguous conditions.
