2. Outline
Introduction
Related Work
Key Cocepts
Existing System
Proposed System
Methodology
Advantages & Disadvantages
Conclusion
2
3. Introduction
Big Data
Service Recommendation
Keyword Aware Service Recommendation(KASR)
Map-Reduce
3
4. Related Work
Item-Based collaborative filtering[1]
Recommendation system based on users history[2]
Bayesian-inference-based recommendation[3]
KASR recommendation[4]
With the development of cloud computing software tools such as
Apache Hadoop, Map-Reduce, it becomes possible to design and
implement scalable recommender systems in “Big Data”
environment.[5]
4
5. Key Concepts
• Recommender System
1. Content-based
2. Collaborative
3. Hybrid recommendation
• Collaborative Filtering
1. Item based
2. User based
• Cloud Computing and MapReduce
5
6. Existing System
6
Nowadays data is increasing explosively so, recommend appropriate
service to user is difficult.
To manage and analyze the big data using the traditional KASR
method is inefficient, it does not uses any big data processing
framework like hadoop.
7. Proposed System
KASR aims at calculating a personalized rating of each candidate
service for a user, and then presenting a personalized service
recommendation list and recommending the most appropriate services to
them.
To improve the scalability and efficiency of our recommendation
method in “Big Data” environment, we have to implement it on a
MapReduce framework on Hadoop by splitting the proposed algorithm
into multiple MapReduce phases.
7
8. Methodology
8
Keyword Aware Service
Recommendation Method
1) Capture user preference by
keyword-aware approach
2) Similarity computation
3) Calculate personalize rating
and generate
recommendation
Implementation on
MapReduce
Fig. KASR main three steps
9. Methodology
9
• Capture user preference
1. Preference of active user
2. Preference of previous user
• Similarity Computation
1. Approximate similarity
computation(ASC)
2. Exact similarity
computation(ESC)
• Calculate personalized ratings
and generate recommendation
Algorithm for SIM-ASC
Input:
The preference keyword set of the
active user APK ,
The preference keyword set of a
previous user PPKj
Output: The similarity of APK and
PPKj, simASC(APK,PPKj)
1. simAPK(APK,PPKj)=
𝐴𝑃𝐾 𝑃𝑃𝐾𝑗
𝐴𝑃𝐾 𝑈 𝑃𝑃𝐾𝑗
2. return the similarity of APK and
PPKj, simASC(APK,UPKj)
11. Implementation
11
Implementation on MapReduce
1. KASR-ASC on MapReduce
Step 1: Process review of candidate service by previous user into their
preference keyword set & compute average rating for each candidate
service.
Step 2: Compute similarity between active user and previous user
Step 3: Calculate personalize rating of each candidate service and
present personalized recommendation list to active user. Based on this
recommendation is obtained
12. Implementation
12
2. KASR-ESC on MapReduce
Step 1: The first step of KASR-ESC on MapReduce is same as step 1 of
KASR-ASC on MapReduce
Step 2: Process all review of each previous user into corresponding
keyword sets respectively.
Step 3: Compute similarity between active and previous user
Step 4: Same as step 3 of KASR-ASC on MapReduce. Based on this
approach present personalized recommendation list to active user and
recommend most appropriate service to him/her.
13. Advantages & Disadvantages
13
• Advantages
i. The proposed method presenting a
personalized service
recommendation list and
recommending the most appropriate
services to the users.
ii. By implementing the KASR in Big
Data environment we have improved
the scalability and efficiency.
iii. The accuracy of the service
recommender systems over exiting
approaches will be improved.
iv. Data analysis will be faster with the
growth of data requirements.
Disadvantages/ Limitations
i. If the specific service has no review
about itself then it difficult to
recommend that service.
ii. There are chances of dummy/fake
reviews.
iii. Some reviews are just for
formalities while descriptive
reviews are required.
14. Conclusion
14
In this seminar we studied KASR which aims at presenting
personalized service recommendation, list and recommending most
appropriate service to user.
To improve scalability and efficiency of KASR in “Big Data”
environment we have to implement it on MapReduce framework in
Hadoop Platform.
15. References
15
[1] Badrul Sarwar, George Karypis, Joseph Konstan, and John
Riedl, ”Item Based Collaborative Filtering Recommendation
Algorithms," ACM, 10, May 15, 2001.
[2] G.Kang, J. Liu, M. Tang, X. Liu and B. Cao, “AWSR:
Active Web Service Recommendation Based on Usage History,"
2012 IEEE 19th International Conference on Web Services
(ICWS), pp. 186-193, 2012.
[3] X. Yang, Y. Guo, Y. Liu, “Bayesian-inference based
recommendation in online social networks," IEEE Transactions
on Parallel and Distributed System,Vol. 24, No. 4, pp. 642-651,
2013.
16. References
16
[4] S. Meng, W. Dou, X. Zhang, J. Chen, “A Keyword-Aware Service
Recommendation based on map reduce," IEEE Transaction on Parallel
and distributed system, DOI 10.1109/TPDS.2013.2297117.
[5] J. Dean, and S. Ghemawat, “MapReduce: Simplied data processing
on large clusters," Communications of the ACM, Vol. 51, No.1, pp.
107-113, 2005.