大量の属性データからの興味深いルールの抽出手法 : 新規分譲マンションの契約者データへの適用

抄録

For newly built condominium supply, it is crucial to uncover characteristics of the potential buyers' behavior as "rules" or "patterns" from historical data. A new approach is applied to extract informative and unexpected rules or patterns from a database with relatively few transactions but a huge number of attributes using data mining methodology. For the val idity of the proposed method, the real data of 798 condominium purchasers in Tokyo metropolitan area are analyzed. Results include that if access time to the nearest railway station to be over 10 minutes, then "accessibility" as a resigned factor increases, which implies that potential accessibility in mind would be around 10 minutes on walk or within about 350 meters from nearest railway station.