17.txt

This complete matlab for neural network

Other Experiments 

To ascertain the utility of the techniques described above, a series of other 
experiments were conducted with data using the site http://acr-news.org which 
has a dedicated server for the newsletter of the Association for Consumer Rese
arch. The site contains a variety of news items, including President's columns
, conference announcements, and call-for-papers for a number of conferences an
d journals. As an example we describe the results of one experiment using tran
saction clustering technique described in this paper. The session file for the
 experiments contained 18430 user transactions with a total of 192 unique URLs
. Again, based on the average session size, a window of size 2 was chosen for 
the experiment. The sample path tested was: 

President's Column => Conference Update => Calls for Papers => 
Special Topics => Note from the Editor =>ACR News Online Archives => 
ACR 2000 Asia-Pacific Conference
A summary of the results is provided in the table below. The recommendations a
re presented using the title tag of HTML pages associated with the URLs in the
 recommendation set. 
  
  
Session Window Recommendation Score 
President's Column
 
 News from the Grapevine 0.39 
ACR News Online Archives 0.37 
Note from the Editor 0.35 
Conference Update 0.34 
Special Topics  0.32 
President's Column 
Conference Update
 
 Calls for Papers 0.52 
ACR 1999 European Conference 0.47 
Note from the Editor 0.43 
1999 ACR Annual Conference 0.39 
ACR 2000 Asia-Pacific Conference 0.39 
ACR News Online Archives 0.34 
ACR News Updates 0.33 
Conference Update 
Calls for Papers
 
 ACR 1999 European Conference 0.72 
ACR News Updates 0.64 
1999 ACR Annual Conference 0.57 
ACR 2000 Asia-Pacific Conference 0.56 
Note from the Editor 0.36 
Journal of Psychology and Marketing 0.35 
Seminar in Marketing Communications 0.34 
ACR News Online Archives 0.34 
Calls for Papers 
Special Topics
 
 ACR News Updates 0.63 
ACR 1999 European Conference 0.58 
1999 ACR Annual Conference 0.57 
Journal of Psychology and Marketing 0.50 
ACR News Online Archives 0.43 
Note from the Editor 0.42 
ACR 2000 Asia-Pacific Conference 0.40 
News from the Grapevine 0.39 
Special Topics 
Note from the Editor 
 
 ACR News Online Archives 0.46 
News from the Grapevine 0.37 
Note from the Editor 
ACR News Online Archives 
 
 Journal of Consumer Psychology 0.41 
President's Column - Sept 1997 0.36 
Journal of Psychology and Marketing 0.36 
ACR News Online Archives 
ACR 2000 Asia-Pacific Conference
 
 1999 ACR Annual Conference 0.76 
Journal of Consumer Psychology 0.41 
Journal of Psychology and Marketing 0.36 
President's Column - Sept 1997 0.34 
Calls for Papers 0.31 




System Implementation and Demonstration Site

The ACR News site discussed in the previous section was used to implement a de
monstration version of the Web personalization system based on the techniques 
and the architecture presented in this paper. A local version of the site whic
h uses our recommendation engine is available for demonstration purposes from 
http://aztec.cs.depaul.edu/scripts/ACR2. For this demonstration, we used a sub
set of the ACR logs (from June 1998 to June 1999), and used transaction cluste
ring to derive URL clusters. The transaction clustering process yielded 28 URL
 clusters representing different types of user access patterns. A threshold of
 0.5 was used to derive URL clusters from transaction clusters (i.e., URL clus
ters contained only those URL references appearing in at least 50% of transact
ions). We used a recommendation threshold of 0.3 as a cut-off point to ensure 
capturing overlapping user interests. Based on the average session size, the s
ystem automatically chose a session window of size of 3 references.

The recommendation engine was implemented as a set of CGI scripts, using cooki
es to keep track of user's active session. Figures below depict a typical inte
raction of user with site. The top frame in each window contains the actual pa
ge contents from the site, while the bottom frame contains the recommended lin
ks. When the user clicks on a link in either frame, the top frame will display
 the content of the requested page, and the bottom frame is dynamically update
d to include the new recommendations. As seen in Figure 3, initially the syste
m does not provide any recommendations until the user has navigated through mo
re pages. Figure 4 shows the recommendations resulting after the user has foll
owed a path to "President's Column" and then to "Online Archives." The recomme
ndations include past President Columns and Editor's Notes (as well as other p
ages) often visited by users who have shown similar access patterns. Figure 5 
and 6 show the results of the user navigation through "Conference Update," "Call for Papers," and then "1999 Asia Pacific Conference". As can
 be seen in these Figures, user's intention of looking for more specific infor
mation will result in more specific recommendations. For example, in Figure 5,
 general recommendations are provided guiding the user to upcoming conferences
 and news items. When the user accesses a specific conference page (Figure 6),
 other specific conference information is presented as potentially interesting
 (e.g., "Winter 2000 SCP Conference" and "Int'l Conference on Marketing and De
velopment").






Figure 3. Main page for the demonstration site. Initially, no recommendations 
are provided as the active user session does not contain sufficient number of 
references. 





Figure 4. Dynamic recommendations after the user has navigated through "Presid
ent's Column" and "Online Archive" pages.
 





Figure 5. Recommendations based on user navigation through "Conference Update"
 and "Call for Papers" pages.
 





Figure 6. System provides specific recommendations related to conferences, bas
ed on user navigation through "Conference Update," "Call for Papers," and "199
9 Asia Pacific Conference" pages.
 



  
Conclusions 

The Web is providing a direct communication medium between the vendors of prod
ucts and services, and their clients. Coupled with the ability to collect deta
iled data at the granularity of individual mouse clicks, this provides a treme
ndous opportunity for personalizing the Web experience for clients. In e-comme
rce parlance this is being termed mass customization. Outside of e-commerce al
so, the idea of Web personalization has many applications. Recently there has 
been an increasing amount of research activity on various aspects of the perso
nalization problem. Most current approaches to personalization by various Web-
based companies rely heavily on human participation to collect profile informa
tion about users. This suffers from the problems of the profile data being sub
jective, as well getting out of date as the user preferences change over time.
 

We have provided several techniques in which the user preference is automatica
lly learned from Web usage data, by using data mining techniques. This has the
 potential of eliminating subjectiveness from profile data as well as keeping 
it up-to-date. We describe a general architecture for automatic Web personaliz
ation based on the proposed techniques, and discuss solutions to the problems 
of usage data preprocessing, usage knowledge extraction, and making recommenda
tions based on the extracted knowledge. Our experimental results indicate that
 the techniques discussed here are promising, each with its own unique charact
eristics, and bear further investigation and development. 
  

References 

[AAP99] Agarwal, R., Aggarwal, C., and Prasad, V., A tree projection algorithm
 for generation of frequent itemsets. In Proceedings of High Performance Data 
Mining Workshop, Puerto Rico, 1999. 

[AS94] Agrawal, R. and Srikant, R., Fast algorithms for mining association rul
es. In Proceedings of the 20th VLDB conference, pp. 487-499, Santiago, Chile, 
1994. 

[BM99] Buchner, A. and Mulvenna, M. D., Discovering internet marketing intelli
gence through online analytical Web usage mining. SIGMOD Record, (4) 27, 1999.
 

[Cha96] Charniak, E., Statistical language learning. MIT Press, 1996. 

[CC99] Clifton, C. and Cooley, R., TopCat: data mining for topic identificatio
n in a text corpus. In Proceedings of the 3rd European Conference of Principle
s and Practice of Knowledge Discovery in Databases, Prague, Czech Republic, 19
99. 

[CMS99] Cooley, R., Mobasher, B., and Srivastava, J., Data preparation for min
ing World Wide Web browsing patterns. Journal of Knowledge and Information Sys
tems, (1) 1, 1999. 

[CPY96] Chen, M. S., Park, J. S., and Yu, P. S., Data mining for path traversa
l patterns in a Web environment. In Proceedings of 16th International Conferen
ce on Distributed Computing Systems, 1996. 

[HBG+99] Han, E-H, Boley, D., Gini, M., Gross, R., Hastings, K., Karypis, G., 
Kumar, V., and Mobasher, B., More, J., Document categorization and query gener
ation on the World Wide Web using WebACE. Journal of Artificial Intelligence R
eview, January 1999. 

[HKBR99] Herlocker, J., Konstan, J., Borchers, A., Riedl, J.. An algorithmic f
ramework for performing collaborative filtering. To appear in Proceedings of t
he 1999 Conference on Research and Development in Information Retrieval, Augus
t 1999. 

[HKKM97] Han, E-H, Karypis, G., Kumar, V., and Mobasher, B., Clustering based 
on association rule hypergraphs. In Proccedings of SIGMOD’97 Workshop on Rese
arch Issues in Data Mining and Knowledge Discovery (DMKD’97), May 1997. 

[HKKM98] Han, E-H, Karypis, G., Kumar, V., and Mobasher, B., Hypergraph based 
clustering in high-dimensional data sets: a summary of results. IEEE Bulletin 
of the Technical Committee on Data Engineering, (21) 1, March 1998. 

[KMM+97] Konstan, J., Miller, B., Maltz, D., Herlocker, J., Gordon, L., and Ri
edl, J., GroupLens: applying collaborative filtering to usenet news. Communica
tions of the ACM (40) 3, 1997. 

[NFJK99] Nasraoui, O., Frigui, H., Joshi, A., Krishnapuram, R., Mining Web acc
ess logs using relational competitive fuzzy clustering. To appear in the Proce
edings of the Eight International Fuzzy Systems Association World Congress, Au
gust 1999. 

[PE98] Perkowitz, M. and Etzioni, O., Adaptive Web sites: automaticlly synthes
izing Web pages. In Proceedings of Fifteenth National Conference on Artificial
 Intelligence, Madison, WI, 1998. 

[SF99] Spiliopoulou, M. and Faulstich, L. C., WUM: A Web Utilization Miner. In
 Proceedings of EDBT Workshop WebDB98, Valencia, Spain, LNCS 1590, Springer Ve
rlag, 1999. 

[SKS98] Schechter, S., Krishnan, M., and Smith, M. D., Using path profiles to 
predict HTTP requests. In Proceedings of 7th International World Wide Web Conf
erence, Brisbane, Australia, 1998. 

[SM95] Shardanand, U., Maes, P., Social information filtering: algorithms for 
automating "word of mouth." In Proceedings of the ACM CHI Conference, 1995. 


[SZAS97] Shahabi, C., Zarkesh, A. M., Adibi, J., and Shah, V., Knowledge disco
very from users Web-page navigation. In Proceedings of Workshop on Research Is
sues in Data Engineering, Birmingham, England, 1997. 

[YJGD96] Yan, T., Jacobsen, M., Garcia-Molina, H., Dayal, U., From user access
 patterns to dynamic hypertext linking. In Proceedings of the 5th Internationa
l World Wide Web Conference, Paris, France, 1996. 

--

※ 来源:．南京大学小百合站 http://bbs.nju.edu.cn [FROM: 202.112.59.146]