METHODOLOGY


Our proposed hybrid recommendation mechanism is composed of four phases, as shown in Figure 2. The first phase is to extract association rules from the web log database. Among the data mining techniques, association rules mining algorithm has been popular in marketing intelligence fields (Lee et al., 2002). Therefore, we applied association rules mining to the web data mining tasks . The web log database, which has been used in data mining, includes the web surfing log files (time, frequency, duration, products, etc.) users made on a target shopping mall or web site. From a data preprocessing viewpoint, the web log data poses the following challenges: (1) large errors, (2) unequal sampling, and (3) missing values. To remove these noises included in data, we applied preprocessing techniques to web log data. Through web data mining, we can usually find the hidden informative relationships between those products and the interrelated hyperlinks users visited while web surfing. Association rules are similar to IF-THEN rules, in which a condition clause (IF) triggers a conclusion clause (THEN). In addition, association rules include the support and confidence (Agrawal et al., 1993a, 1993b). The association rules mining algorithm is shown in Table 1.

click to expand
Figure 2: Research Methodology of Hybrid Recommendation
 
Table 1: Pseudo Code of the Association Rules Mining Algorithm

C k : Candidate transaction set of size k

L k : Frequency transaction set of size k

L j = {frequent items};

For ( k =1; L k != ˜; k ++) Do Begin

C k+1 = Candidates generated from L k;

For Each transaction t in database Do

       Increment the count of all candidates in C k+1 that are contained in tL k+1 = candidates in C k+1 with min_support

End Return L k ;

In the second phase, after the extraction of the association rules, we adapt CBR to extend the quality of reasoning and recover the limitation of rule-based reasoning. CBR is both a paradigm for computer-based problem-solvers and a model of human cognition. Therefore, cases extracted from the customer database may imply the customer's knowledge of products and predict his future behavior. Through this phase, CBR shows significant promise for improving the effectiveness of complex and unstructured decision-making.

The third phase is to build a hybrid knowledge base. In this phase, we combine rule base with case base. The key features to combining these two different knowledge bases are the customer's profile and the products.

The final phase of the proposed hybrid recommendation mechanism is to apply inference procedures to the hybrid knowledge base and extract the inference results. Figure 2 shows our proposed mechanism.




(ed.) Intelligent Agents for Data Mining and Information Retrieval
(ed.) Intelligent Agents for Data Mining and Information Retrieval
ISBN: N/A
EAN: N/A
Year: 2004
Pages: 171

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net