Systems, Man, and Cybernetics

 

Home

• [P350] Visualizing Method based on Item Sales Records and its Experimentation
  Author(s): Yoshihiro Hayashi, Hiroshi Tsuji, Ryosuke Saga
  
  

  Abstract
  As transaction logs are accumulated in database, how to use them for making strategy is an attractive problem. Previously for the commodity items such as beer and soap, a data mining method based on the preference transition was proposed. To examine the method, this paper describes prospects of sensitivity analyses by changing the threshold of the transition strength, which is expressed as Simpson coefficient. The sensitivity on visualization is also discussed by deleting special biased nodes. A numerical example of the sales records of 172 kinds of products and their 90,000 sales records is also illustrated.

  
   

Full Text (PDF)