Professional Documents
Culture Documents
User Behavior Analysis Based On User Interest by Web Log Mining
User Behavior Analysis Based On User Interest by Web Log Mining
Abstract: With the rapid development of science and introduces the user’s browsing behavior model
technology and the growing popularity of computer construction. Section 5 introduces the M5 model based on
networks, the scale of network users is gradually expanding, the user behavior analysis method. Chapter 6 carries on the
and the behavior of network users is becoming more and experimental analysis to the model. Chapter 7 concludes
more complicated. A large number of studies show that the the paper.
user’s actual interest is closely related to the browsing
behavior on the web page. Through the user browsing II. RELATED WORK
behavior analysis can obtain the user interest information,
and then build the user interest model, so that the search
The traditional data mining technology and web
results closer to the user's expectations. This paper mainly integration for web mining. Web digging is the extraction
introduces the method of web log mining, which can of interesting, potentially useful patterns and hidden
discover the mode of web pages by digging web log records. information from web documents and web activities. Web
By analyzing and exploring the rules of web log records, we data mining is the use of data mining technology,
can identify the potential customers of the website and automatically discover and extract information from web
improve the quality of information services to users. In the documents and services.. In general, according to the
stage of user behavior analysis, this paper explores the different data mining object, web data mining is divided
differences in user browsing behavior in different types of into three types: web content mining, web structure
access events, and calculates the user's interest based on the mining, web log mining. [9].
M5 model tree to analyze the analytic events.
User interest modeling is the process of summing up a
Keywords: network user behavior; data mining; user interest; computable user interest model from information that can
M5 model tree reflect user preferences. Researchers have done a lot of
work in the area of user interest modeling and have made
I. INTRODUCTION a lot of valuable results. MariamDaoud and other
With the rapid development of the Internet to researchers, based on the ontology, proposed a user
information access, communication and communication- interest modeling method based on semantic graph [17].
based basic network services are gradually developed for Hochd Jeon proposed a method of using adaptive updating
the leisure and entertainment, electronic services, e- strategy for adaptive user interest modeling [18].
commerce services to expand the three major categories of Adomavicius and Tuzhilin used the data mining method
network services. Users in the visit of the page and the to mine the access records of the individual users, excavate
user's interest is closely related, such as the care of the association rules and the user registration of the
financial users will often visit some financial class of the personal information constitute the user model. Sofia
site, and like sports users will often visit some sports news Stamou and Alexandros Ntoulas proposed a method of
sites or sporting goods website. We can use the user's visit user interest modeling by analyzing query terms and web
records to tap the user's interest in a topic. page subject information. Researchers such as Paul
conducted user interest modeling through ODP
The main purpose of the study of the user's visit record classification system and data information.
is to analyze the user's most concerned about the results
from the mining results. By analyzing the user access to The M5 model tree is proposed by Quinlan to solve the
resources of the time, frequency and so on,modify the problem of continuous value prediction. The M5 model
structure and design of the site to expect more customers tree is a decision tree that uses linear regression function
to stay and better serve customers. User behavior analysis at leaf nodes. By using a standard method to convert the
has become a new research hotspot. The work of this paper classification problem into function optimization problem,
mainly studies the data mining technology in user the final model can be expressed as a piecewise linear
behavior analysis, and builds the user interest model based function. M5 algorithm can be divided into model tree
on the user interest information, and finally draws the construction and linear regression model
user's interest. The above theory and model algorithm help us to use
The rest of this paper is organized as follows. Section web log mining technology to obtain user access behavior
2 describes the relevant work of the paper. Section 3 data, and on this basis, based on the M5 model tree to
introduces the Web log mining technology. Section 4 construct the user behavior model for user interest.
C Analysis of results
Calculated page interest, the model can be evaluated
and the accuracy of the page is
Fig. 2 General Prediction Mode
2017 27th International Telecommunication Networks and Applications Conference (ITNAC)
,"49.7.89L"49.7.89 [1] Wan Fei, Zhao Xi, Liang, and so on. Research on Search Engine
%+/.%7, M L 7
,"49.7.89 User Behavior Based on Mobile Internet Log. Chinese Journal
of Information, 2014,28 (2): 144-150.
Where PagePrc is the accuracy of the page, Aclnterest [2] Cen Rongwei, Liu Yiqun, Zhang Min, et al. Search Engine User
is the user's actual interest in the page, determined by the Behavior Analysis Based on Log Mining. Chinese Journal of
user's manual score. Interest is the user's interest analysis Information, 2010,24 (3): 49-54.
model to infer the user's interest in the page. The accuracy [3] Rong Guoting, Luo Yong, Sun Jianjun. Research on Library
of the user behavior model for a particular user is the User 's Behavior Based on Log Analysis. Library Journal, 2015
(7): 59-63.
average of the user's access to the page accuracy. [4] N. Ghahreman and M. Sameti, “Comparison of M5 model tree
A H and Artificial Neural Network for estimating Potential
M GF@
(8) Evapotranspiration in semi- arid climates,” Department of
H
Irrigation and Reclamation Engineering, University of Tehran,
Where Accuracy represents the accuracy of the model Karaj, Iran, March 2014.
to the user and n is the number of pages visited by the user. [5] P. Ditthakit and C. Chinnarasri, “Estimation of Pan Coefficient
using M5 Model Tree”, School of Engineering and Resources,
This article will ask the user to mark interest after the Walailak University, Nakhon Si Thammarat 80160, Thailand,
user has finished browsing. Interest degree values include 2012.
0 ~ 3 five grades. Again, according to the interest rate [6] Yan, Q.,Wu, L.,Zheng, L. et al.Social network based microblog
estimation method to calculate the page interest rate, and user behavior analysis[J].Physica, A. Statistical mechanics and
its applications,2013,392(7):1712-1723.
compared with the results of user labeling, get the [7] Zhenhua Wang,Lai Tu,Zhe Guo et al.Analysis of user behaviors
accuracy of the model on the page. Finally, the accuracy by mining large network data sets[J].Future generations
of the model for each user is calculated from equation (7) computer systems: FGCS,2014,37:429-437.
and (8). [8] Yun Liu,Weiguo Yuan.User Posting Behavior Analysis and
Modeling in Microblog[C].//2014 Tenth International
Based on all the behavior data of each user collected, Conference on Intelligent Information Hiding and Multimedia
the general forecasting model of all the data is constructed Signal Processing: 2014 Tenth International Conference on
based on the M5 model tree. The accuracy rate of the Intelligent Information Hiding and Multimedia Signal
Processing (IIH-MSP 2014), 27-29 August 2014, Kitakyushu,
corresponding model is obtained according to the formula Japan.2014:916-919.
(7) and (8)The accuracy rate is shown in Figure 5. [9] Yin B, Zhang Z, Wang X, et al. Research and Application of
Data Mining Technology Used in the Analysis of Smart Home
User Behavior[C]// Sixth International Conference on
Measuring Technology and Mechatronics Automation. IEEE,
2014:476-479.
[10] Hájek P, Stejskal J. Library user behavior analysis - Use in
economics and management[J]. Wseas Transactions on Business
& Economics, 2014, 11(1):107-116.
[11] Jaewon Kim,Paul Thomas,Ramesh Sankaranarayana et al.Eye-
Tracking Analysis of User Behavior and Performance in Web
Search on Large and Small Screens[J].Journal of the Association
for Information Science and Technology,2015,66(3):526-544.
[12] Long Chen,Yong-Qing Wang.Forensic Analysis towards the
user behavior of Sina microblog[C].//International conference
on education technology, management and humanities science:
ETMHS 2015, Xi an, China, 21-22 March 2015, Part 2 of
Fig. 5 The accuracy results 2.2015:1167-1171.
[13] Mi Zhang,Christopher C. Yang.Using Content and Network
As can be seen from Figure 5, average for all users' Analysis to Understand the Social Support Exchange Patterns
accuracythe accuracy of the general forecasting model and User Behaviors of an Online Smoking Cessation
is 65.2%, so the user behavior analysis method based on Intervention Program[J].Journal of the Association for
the M5 model can be applied to the prediction of the user's Information Science and Technology,2015,66(3):564-575.
[14] Behnood, Ali,Olek, Jan,Glinicki, Michal A. et al.Predicting
interest to a certain extent. modulus elasticity of recycled aggregate concrete using M5 '
model tree algorithm[J].Construction and Building
VII. CONCLUSION Materials,2015,94(Sep.30):137-147.
User behavior analysis is through the way of data [15] Nitha Ayinippully Nalarajan,C. Mohandas.Groundwater Level
Prediction using M5 Model Trees[J].Journal of The Institution
mining from a large number of network information
of Engineers (India), Series A. Civil, architectural,
mining user behavior patterns. It is a relatively new environmental and agricultural engineering,2015,96(1):57-62.
research field, has a wide range of application prospects, [16] JIA Ming-ming.New Method for Generating Model Tree [J].
become a hot topic of domestic and foreign scholars. Software Journal, 2008 (04): 35-37.Teevan, J.,Dumais,S.T., and
Horvitz,E. (2010). Potential for personalization. ACM.Interact.
This paper mainly studies the behavior analysis 17,1,1-31.
method of user interest, studies the Web log mining [17] Jansen B J, Booth D L,Spink A, Determining the informational,
method, and puts forward the calculation degree of user's navigational, andtransactional intent of Web queries [J].
interest in web pages based on M5 model. The process of Information Processing & Management, 2008,44(3): 1251-1266.
[18] Adomavicius G and Tuzhilin A. Using Data Mining Methods to
constructing user behavior model based on M5 is Build Customer Profiles.IEEE Computer. Feb 2001:74-82.
introduced in detail. Finally, we use the collected user
behavior data to construct a general model of user interest
forecast.