PSO Based User Grouping Case Study

4.1. PSO Based User Grouping
Grouping of users is a critical assignment in web usage mining procedure. For user clustering,
PSO methodology is connected. Particle Swarm Optimization (PSO) is a SI (Swarm Intelligence) based advancement method which was created by Kennedy and Eberhart in the mid 1990s (Kennedy et al., 1995). The optimization method of Particle Swarm Optimization (PSO) reproduces the development and movement of animals. In this procedure, every user is alluded as particle and it is introduced with a situated of 50 users arbitrarily in the searching space. Here the users goes with a same rate of interest change are considered for clustering. The average rate of changes in interest is assessed by utilizing the cosine similarity
This worth ought to be somewhere around 0.9 and 1.2 to enhance the PSO’s execution. d1 and d2 are two consistently circulated random variables in the extent [0,1] and d1=d2=1. a1 and a2 are certain consistent parameters called speeding up coefficients which controls the most extreme step size between progressive cycles. For the values of a1=a2=2 PSO produces the most upgraded grouping result.
Interest change rate and class of every user are found at every cycle by utilizing the above equations (Ahmadyfard et al., 2008). These values can be changed in progressive cycles. The user may change to another cluster by the estimation of rate of interest change as per the above equation. The stop condition takes the greatest number of cycles. It gives the most improved result for user grouping. The PSO approach for web user grouping can be given in Figure 3.
User grouping is a procedure of categorizing users into an arrangement of clusters with their similarity in web search. For this reason, PSO approach is used and Figure 4 demonstrates
Education, Sports, Engineering, Computer, Entertainment, News, Travel, Finance, Music and Others can be taken as classes. User 0 and User-31 goes under the classification of Education.
Also, different users are clustered under their corresponding groups (Sumathi et al., 2014).
4.2. New User Categorization
Categorization is the process of finding the category for new user based on the search concept or new user’s query. The purpose of this categorization is to retrieve the similar user web pages for ranking. The proposed ranking approach analyzes the existing users’ interests to recognize commonalities and rank the pages which is relating to the preferences from other users with similar tastes. The new user is categorized based on the existing users searching behavior. For
Copyright © 2015, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
26 International Journal of Intelligent Information Technologies, 11(4), 16-36, October-December