Application of salesman-like recommendation system in 3G mobile phone online shopping decision support

  • Authors:
  • Ching-Torng Lin;Wei-Chiang Hong;Yi-Fun Chen;Yucheng Dong

  • Affiliations:
  • Department of Information Management, Da-Yeh University, 168 University Rd., Dacun, Changhua 51591, Taiwan;Department of Information Management, Oriental Institute of Technology, 58 Sec. 2, Si-Chuan Rd., Panchiao, Taipei 220, Taiwan;Department of Information Management, Da-Yeh University, 168 University Rd., Dacun, Changhua 51591, Taiwan;Department of Management Science, School of Management, Xi'an Jiaotong University, Xi'an 710049, PR China

  • Venue:
  • Expert Systems with Applications: An International Journal
  • Year:
  • 2010

Quantified Score

Hi-index 12.05

Visualization

Abstract

The rapid growth of e-commerce has confronted both enterprises and consumers with a new situation. Whereas companies are finding it harder to survive, consumers are unable to effectively select the products that really to meet their needs. To reduce the product overload of Internet shoppers, a variety of recommendation techniques that track previous actions of groups of consumers to make personalized recommendations have been developed and applied. Current personalized recommendation systems suffer from the need to analyze large sets of consumer data, or data for numerous consumers. However, even within a single group, consumer preferences may differ, and individual preferences may also change with circumstances. Additionally, the consumer product knowledge influences their browsing actions. To orient Web-visitors on how to become consumers, a salesman-like recommendation technology was developed based on visitor product preference index, which comprises of their product knowledge and browsing actions in scene. A prototype system for use with high-technology product, 3G phones, was developed to test the effectiveness of the recommendation technology. Through a test of 250 objectives, the results show that the recommendation deviation level can be reduced to 0.49, and exact fit with visitor favor products can reach 60.8%, showing that the proposed model can achieve recommendation effectiveness.