Online Recommendation System
Online Recommendation System
Online Recommendation System
SJSU ScholarWorks
Master's Projects Master's Theses and Graduate Research
2008
Recommended Citation
Khera, Ankit, "Online Recommendation System" (2008). Master's Projects. 97.
DOI: https://doi.org/10.31979/etd.v33b-ap2s
https://scholarworks.sjsu.edu/etd_projects/97
This Master's Project is brought to you for free and open access by the Master's Theses and Graduate Research at SJSU ScholarWorks. It has been
accepted for inclusion in Master's Projects by an authorized administrator of SJSU ScholarWorks. For more information, please contact
scholarworks@sjsu.edu.
Online Recommendation System
A Master's Project
Presented to
In Partial Fulfillment
Master of Science
by
Fall 2008
1
© 2008
2
APPROVED FOR THE DEPARTMENT OF COMPUTER SCIENCE
_____________________________________________________
_____________________________________________________
_____________________________________________________
______________________________________________________
3
ACKNOWLEDGEMENT
4
ABSTRACT
The vast amount of data available on the Internet has led to the
and tools used so far. The report also includes the project schedule
and deliverables.
5
TABLE OF CONTENTS
1. Introduction .................................................... 8
4. Implementation: ................................................ 22
7. Conclusion ..................................................... 39
References ........................................................ 40
6
LIST OF FIGURES
7
1. Introduction
on the World Wide Web. Information on the Internet grows rapidly and
results along with it. The system makes use of numerical ratings of
similar items between the active user and other users of the system
The results show that the system rests in its assumption that active
8
users will always react constructively to items rated highly by
summarizing user queries, and linking the metadata like tags and
System would benefit those users who have to scroll through pages of
9
2. Project Overview
Web Browser
Safari/Firefox etc
Knowledge Base
Description:
10
3. The user chooses from amongst the type 2 different types of
recommendation systems available.
and then recommends items to the users based on the most similar
user.
make predictions.
1.1. Auto search complete: The System provides its users with
11
benefits; users can search for books matching authors,
author so far.
2. Rate Books: Users can rate the movies which they like/dislike by
also allows the users to tag their books, and provide feedback.
3. View/Edit past books: The system allows the users to view and edit
12
3. Recommendation Systems:
in.
3.2. Methodologies
13
Finder to implement Context based filtering techniques to generate
14
Figure 2: Taste architecture (Sean, 2008)
15
Tags (free tagging)
Contex
t
Tag Kid's
s Movie
Figure 4: Movie rating parameters
reduce the three ratings for any movie to a single movie rating
m1 m2 m3 m4
users b
movies
16
logic to it and generate a user similarity matrix as shown in
the figure:
Users
Users
The above figure shows the user similarity matrix in which the
condition example: (ab=0.8) > (α=0.4) so user 'a' and user 'b'
17
Figure 7: Similar users(Klir, 1988)
contextual information.
18
users based on the numerical ratings of common items rated by
the active users and other users of the system. The system
give high scores than user ‘B’ but both tend to like the
consistent.
Algorithm:
user ‘A’ and user ‘B’. It then finds out the sums and sum
19
Figure 8: Pearson’s Correlation formula.
good but it did not meet the goals set for the context-based
engine initially. The system did not give good results due to
the CF based engine, the system did not do justice to the word
The system does not expect the user to provide the complete
20
author, ISBN; publisher name example ‘oxf’ could be typed
to type any free keywords. Once the user clicks the submit
and Synonym Finder (senses) are then shown to the user. The
users.
21
4. Implementation:
1) Home Screen
This is how the home screen for the online recommendation system
22
enter the ‘userID’. We can see this in the above figure were User
‘23446’ has just logged. The session for this user has to remain
active through out the recommendation process in order for the system
to make recommendations.
2) Books Search
The above figure shows the implementation of the auto search feature
23
as described above, the figure displays 10 books with their average
ratings along side matching the keyword ‘ame’ entered by the active
user. If the match is not seen the more link can be clicked to see
3) `More` Keyword
The above figure shows the results of top books matching the keyword
24
4) Results Books Search
The above figure shows the details of the book like isbn, title,
etc. The user can rate the new book or update his current
ratings here.
25
5) Advance Search Books (publisher): -
26
This feature provides the user with advance search capabilities.
The user can search under categories author, ISBN, publisher. The
7) Recommendation
The above figure shows the initial screen shown to the user where
the context information is gathered from the user. The active user
interested in.
27
The above figure shows the collaborative filtering
28
The above figure shows the first set of results shown to the users
In the figure below final results of the context based engine are
29
30
4.2. Technical Specifications
31
4.4. Web Services
Allows the user to tap into the Yahoo! Search technologies from
users. This would help to improve the results of context based engine
1) The php script accepts the keyword 'Madonna' and queries that
keyword to Yahoo Web Service, which returns the results of the
32
Screen Scrapping
the user. This would help the system to find output-improved results.
http://thesaurus.reference.com/
http://www.get-synonym.com/
33
2) The purpose of this php script is to screen scrap synonyms from
a website and use it for recommendations. The script captures
34
3) Result
4.5. Testing
The system has been tested by keeping a small set of data from the
BX-Crossing dataset aside and then monitor whether the system is able
aside database. The system was also tested to see whether the results
ratings.
35
5. Advantages of the System
1) The System would benefit those users who have to use search
engines to locate relevant content. They have to scroll through
2) Rather than searching for quality web pages, the users of this
system would be directly taken to quality web pages matching
deceiving at times.
36
6. Project Schedule/ Deliverables
Schedule
workflow design.
Recommendation
System.
ontology/Taxonomy, extracting
distance scores
between items.
37
from user queries and feedback.
c) Developing an algorithm to
d) Integrating/Implementing the
Collaborative-filtering engine
findings in cs297
planned system.
Deliverables
38
7. Conclusion
context based engine. The system can be highly improved by making use
the speed of the system, using yahoo term extraction web service to
parse and get important keywords from the feeback provided by the
39
References
40
Kumar P, Gopalan S, Sridhar V (2005). Context enabled multi-CBR based
recommendation engine for e-commerce. IEEE International
Conference on e-Business Engineering, 237-255.
http://developer.yahoo.com/
41