Application of Decision-Tree Induction Techniques to Personalized Advertisements on Internet Storefronts

TLDR

Customization and personalization are critical success factors for Internet stores, yet extracting and maintaining rule‑based marketing rules from experts is challenging, despite the underlying theories of statistics, data mining, AI, and rule‑based matching. This paper studies personalized recommendation techniques that suggest products or services to customers of Internet storefronts based on demographics or past purchasing behavior. The authors propose a decision‑tree induction–based rule‑extraction method that generates marketing rules linking customer demographics to product categories and evaluate its effectiveness via preference scoring and random‑selection experiments. The extracted rules provide personalized advertisement selection when a customer visits an Internet store. Keywords: decision‑tree induction, internet advertising, internet storefront, machine learning, personalization.

Abstract

Abstract Customization and personalization services are a critical success factor for Internet stores and Web service providers. This paper studies personalized recommendation techniques that suggest products or services to the customers of Internet storefronts based on their demographics or past purchasing behavior. The underlining theories of recommendation techniques are statistics, data mining, artificial intelligence, and rule-based matching. In the rule-based approach to personalized recommendation, marketing rules for personalization are usually obtained from marketing experts and used to perform inferencing based on customer data. However, it is difficult to extract marketing rules from marketing experts, and to validate and maintain the constructed knowledge base. This paper proposes a marketing rule-extraction technique for personalized recommendation on Internet storefronts using machine learning techniques, and especially decisiontree induction techniques. Using tree induction techniques, data-mining tools can generate marketing rules that match customer demographics to product categories. The extracted rules provide personalized advertisement selection when a customer visits an Internet store. An experiment is performed to evaluate the effectiveness of the proposed approach with preference scoring and random selection. Keywords: DECISION-TREE INDUCTION INTERNET ADVERTISING INTERNET STOREFRONT MACHINE LEARNING PERSONALIZATION