Fr From a a Web eb Ser ervic vices es Catalo alog to a a Li Linked Ecosystem of f Se Services F. SLAIMI, S. SELLAMI, O.BOUCELMA AIX-MARSEILLE UNIVERSITY, FRANCE 1
Outline o Context and motivation o Related work o Graph construction o Recommendation process o Evaluation o Conclusion and future work 2
Outline o Context and motivation o Related work o Graph construction o Recommendation process o Evaluation o Conclusion and future work 3
Context and Motivation Track Increasing number of web services and mashups Track (> 18 K APIs @Pweb) Track L Manual search of services and Track mashups is difficult Track L Services ans mashups are sparse à Tedious process of discovery and recommendation link services/ mashups and users à discovery and recommendation 4
Outline o Context and motivation o Related work o Graph construction o Recommendation process o Evaluation o Conclusion and future work 5
Discovery and recommendation Approaches Graphs Discovery Criteria Selection and recommendation Operates on services, mashups, categories User profiles and preferences of APIs [Guo 2015] and social links between developers Linked Social Services Based on Linked Data Principles Social links [Maamar 2011] Trust based Based on common usage in mashups or by Trust [Deng 2014] [Deng 2015] users QoS evaluations link between mashups of resources which is Similarity Linked mashups calculated based on the comparison of their [Bianchini 2014] terminological items 6
Discovery and recommendation à Most recommendations are based on common usages of services/mashups or result in “same” QoS properties L Ignore services’ properties and mashups (documentation, functional and non functional) L Ignore services and mashups’ similarities L QoS are not always available 7
Outline o Context and motivation o Related work o Graph construction o Recommendation process o Evaluation o Conclusion and future work 8
Services relationships Similarities between categories, names, description and tags Social Facebook Twitter Category : social Category : social Name : facebook Name : twitter Tags : social, Tags : social , 0.72 webhooks microblogging megaphone summary : Social summary : Socia l facebook networking microblogging twitter 0.72 0.7 0.6 fonolo 9 linkedin simS(facebook, twitter)=0,72 twitter facebok
Mashups relationships How to create links between mashups? à common services in mashups |"#$⋂"#&| Sim Mashups (M1,M2) = |"#$∪"#&| facebook facebook Sim Mashups (M1,M2)= & ( = 0,4 M1 M2 filckr LinkedIn Google Google maps maps SMS 10
Users relationships How to create links between users? Links may have different semantics: follows, similarity S1 Similar interests u3 u1 S2 S3 u3 u4 |- ./ ⋂|- .0 | u2 S4 Sim ( u i , u j )= u4 |- ./ | S5 Track relation Where H u i and H u j are the recent histories of users u i and u j respectively 11
Global graph …. Users M1 Mn M2 Mashups u5 u1 …. Services Sn S2 S3 S1 u6 u4 u2 u3 u7 Similar Categories Belongs to …. C1 C2 C3 Cn Track 12
Outline o Context and motivation o Related work o Graph construction o Recommendation process o Evaluation o Conclusion and future work 13
Recommendation process U2 C1 C2 C1 C2 U5 U1 S1 M1 S5 S6 S4 M6 M2 U6 S2 M2 M7 S3 U3 Un Sn Services relationships Users relationships Mashups relationships List of ranked Services/mashups Services services and recommendation Discovery mashups Sub graphs of services User Watchlist U1 User Request 14
Recommendation process Recommended services Services can be used Service’s Category: social with facebook Name: facebook search User A User’s query Recommended mashups 15
Outline o Context and motivation o Related work o Graph construction o Recommendation process o Evaluation o Conclusion and future work 16
System architecture 17
Data set Number of categories 116 Number of mashups 300 Number of services 700 Number of users (with wtachlists) 344 18
Evaluation numbers 0,8 0,9 0,7 0,8 0,6 0,7 0,5 0,6 0,4 0,5 0,3 0,4 0,2 0,3 0,1 0,2 0,1 0 Top 5 Top 10 0 Top 5 Top 10 Precision Recall RMSE hit-rank Precision Recall RMSE hit-rank Recall, Precision, RMSE and Hit-rank numbers (w.r.t the number of recommended Recall, Precision, RMSE and Hit-rank numbers (w.r.t the number of recommended services) mashups) à it is relatively easier to recommend a subset of relevant services. 19
Evaluations numbers o TrsutSVD: Trust based recommendation o Popular: Recommendation of popular services (ProgrammableWeb) Approaches Precision Precision @10 Recall @5 Recall @10 RMSE RMSE @10 @5 @5 TrustSVD 0.73 0.75 0.61 0.63 0.211 0.2 WReG 0.80 0.85 0.70 0.74 0.2 0.185 Popular 0.41 0.39 0.34 0.61 0.31 0.3 20
Evaluation numbers o WReG is based on users-services and users-mashups relationships à recommendations are more precise. o TrustSVD considers trust relations between users and services à gives good precision values à Not able to recommend services in absence of rating values o Popular à lowest results compared to TrustSVD and WReG à does not take into account users’ interests (results are not personalized). 21
Outline o Context and motivation o Related work o Graph construction o Recommendation process o Evaluations results o Conclusion and future work 22
Conclusion o A new web services ecosystem catalog o Multigraph o service à service relations o user à user relations o user à service relations o Neo4J prototype o Recommendation process o Search o Recommendation 23
Future work o Exploit the graph and links between services and mashups to assist the mashups construction process o Extend this work to service management for IoT in order to perform IoT services discovery. 24
References [Maamar 2011] Maamar, Z., Wives, L. K., Badr, Y., Elnaffar, S., Boukadi, K., Faci, N.: Linkedws: A novel web services discovery model based on the metaphor of Social networks. Simulation Modelling Practice and Theory, vol.19 (2011) 121-132 [Guo 2015] Guo, G., Zhang, J., Yorke-Smith., N.: TrustSVD: Collaborative filtering with both the explicit and implicit influence of user trust and of item ratings. Proceedings of the Twenty-Ninth AAAI Conference on Artificial Intelligence, (2015) 123-129 [Deng 2015] Deng, S., Huang, L., Yin, Y., Tang, W.: Trust-based service recommendation in social network. Appl. Math, vol.9 (2015) 1567-1574 [Deng 2014] Deng, S., Huang, L. Xu, G.: Social network-based service recommendation with trust enhancement. Expert Systems with Applications. vol.(41) (2014) 8075-8084 [Bianchini 2014] Bianchini,D., Antonellis, V. D., Melchiori, M.: Link-Based Viewing of Multiple Web API Repositories. In Database and Expert Systems Applications - 25th International Conference, DEXA 2014, Munich, Germany, (2014) 362–376 25
Recommend
More recommend
