Your SlideShare is downloading. ×
A recommender system for social learning platforms
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×

Saving this for later?

Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime - even offline.

Text the download link to your phone

Standard text messaging rates apply

A recommender system for social learning platforms

333
views

Published on

Presentation on #recsys @CWInl

Presentation on #recsys @CWInl

Published in: Technology, Education

0 Comments
3 Likes
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total Views
333
On Slideshare
0
From Embeds
0
Number of Embeds
2
Actions
Shares
0
Downloads
14
Comments
0
Likes
3
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide

Transcript

  • 1. page 1
  • 2. page 2
  • 3. Recommender systems? page 3
  • 4. Open Discovery Space (ODS) A socially-powered, multilingual open learning infrastructure in Europe Recommendations! Which algorithm best fits ODS platform? page 4
  • 5. Towards a Trust-based Recommender for Social platforms Soude Fazeli, PhD candidate Dr. Hendrik Drachsler Prof. Dr. Peter Sloep page 5
  • 6. Recommender algorithms 1. Content-based page 6 2. Collaborative filtering ✓
  • 7. Collaborative filtering algorithms page 7
  • 8. Similarity Sparsity! page 8
  • 9. Trustworthy users == like-minded users Improving prediction accuracy of recommendations page 9
  • 10. •  •  •  •  •  Golbeck’s TidalTrust Trust-aware recommender by Massa and Avesani Andersen et al’s axiomatic approach T-BAR by Bellaachia and Alathel And many more… All require users to give explicit trust ratings! page 10
  • 11. •  Lathia et al.’s trust-based recommender #neal-lathia #recsys •  Trust model by O’Donovan and Smyth Centralized approaches à scalability problem page 11
  • 12. A social recommender system: T-index approach 1. Description •  Trust networks: a graph •  Nodes: users •  Edges: trust relationships •  Weights: trust values originating from similarity •  Each user can be assumed as an agent •  Improve the process of finding nearest neighbors •  T-index •  TopTrustee page 12
  • 13. A social recommender system: T-index approach 2. Trust propagation mechanism •  A new trust relationship between two far unconnected users is inferred if and only if: •  Condition 1: •  page 13 •  Mutual trust value between intermediate users is higher than a certain threshold (v) Condition 2: •  The number of connecting edges is lower than an upper bound (L)
  • 14. rated rated Alice Bob Carol rated rated if A trusts B and B trusts C, then A trusts C if and only if condition 1 is met and condition 2 is met page 14
  • 15. A social recommender system: T-index approach 3. T-index? •  T-index: measure of users’ trustworthiness •  H-index: the impact of publications of an author Indegree (ua) = 7 Indegree (ub) = 5 T-index (ua ) = 2 T-index (ub ) = 4 page 15 Note! Cluster: a group of users who all trust a common user as the most trustworthy one (central user)
  • 16. A social recommender system: T-index approach 3. T-index? page 16
  • 17. A social recommender system: T-index approach 4. What T-index is for? •  TopTrustee : a list of top raters of an item sorted by T-index •  Helps the process of finding nearest neighbors •  page 17 Providing access to trustworthy users across the trust network including even those outside the traversal path length limit (L)
  • 18. Social data page 18
  • 19. •  RQ1: How to generate more accurate and thus, more relevant recommendations by using the social data originating from social activities of users within an online environment? •  RQ2: Can the use of the inter-user trust relationships that originate from the social activities of users within an online environment further evolve the network of users? page 19
  • 20. Hypothesis page 20
  • 21. •  H1: The extended T-index algorithm outperforms the classical collaborative filtering algorithms in terms of F1 score, which is a combined value of precision and recall of the generated recommendations. •  H2: Using T-index-based trust relationships between users improves the structure of the trust network by providing balanced degree distribution of the users. page 21
  • 22. Data-driven study 1. Goal To find out which recommender algorithms best performs and thus, is suitable for social online platforms like ODS platform page 22
  • 23. Data-driven study 2. Method •  Testing several recommender algorithms •  Classical collaborative filtering algorithms using traditional nearest neighbors method •  T-index approach using graph-based method •  Datasets •  MovieLens – standard dataset •  MACE, OpenScout, Travel well -- similar to the future ODS dataset •  Using Mahout page 23
  • 24. Data-driven study 3. Setting •  v = 0.1 (Condition 1), L = 2 (Condition 2) •  Training set 80% and test set 20% •  Sizes of neighborhoods n= (3,5,7,10) •  Size of TopTrustee list m=5 page 24
  • 25. Data-driven study 4. Result 4.1. F1 score 0.1" 0.09" 0.08" 0.07" 0.06" 0.05" 0.04" 0.03" 0.02" 0.01" 0" OpenScout% 0.14" 0.12" 0.1" 0.08" Tanimoto3Jaccard"(CF1)" 0.06" Loglikelihood"(CF2)" Euclidean"(CF3)" 0.04" Euclidean"(CF3)" Graph4based"(CF4)" 0.02" Graph3based"(CF4)" Tanimoto4Jaccard"(CF1)" Loglikelihood"(CF2)" F1@10% F1@10% MACE% 0" 3" 5" 7" 3" 10" 5" Travel%well% 10" MovieLens% 0.25" 0.08" 0.2" 0.06" Tanimoto3Jaccard"(CF1)" 0.04" Loglikelihood"(CF2)" Euclidean"(CF3)" 0.02" Graph3based"(CF4)" 0" 3" 5" 7" size%of%neighborhood%(n)% page 25 10" F1@10% 0.1" F1@10% 7" size%of%neighborhood%(n)% size%of%neighborhood%(n)% Tanimoto0Jaccard"(CF1)" 0.15" Loglikelihood"(CF2)" 0.1" Euclidean"(CF3)" 0.05" Graph0based"(CF4)" 0" 3" 5" 7" 10" size%of%neighborhood%(n)% F1 of the extended T-index and Tanimoto algorithms for different datasets, based on the size of neighborhood
  • 26. Data-driven study 4.2. Created trust network Without T-index page 26 With T-index
  • 27. 2. Data-driven study 4.3. Degree centrality 250" 200" 150" MovieLens" degree% OpenScout" 100" MACE" Travel"well" 50" 0" u1" u2" u3" u4" u5" u6" u7" u8" u9" u10" Top)10%central%users% Degree distribution of top-10 central users for different datasets page 27
  • 28. Conclusion •  The aim is to support user in social platforms to find the most suitable content or people •  Recommender systems can be a solution •  Using trust-based approaches to improve performance of recommender systems even in case of sparsity page 28
  • 29. Ongoing and Further work •  Testing recommender algorithms on more datasets coming from social networking sites •  Go online with the ODS platform (October 2013) •  User evaluation study (February 2013) page 29
  • 30. Soude  Fazeli   PhD  candidate   Open  University  of  the  Netherlands   Centre  for  Learning  Sciences  and  Technologies   (CELSTEC)  PO-­‐Box  2960   6401  DL  Heerlen,  The  Netherlands   email:  soude.fazeli@ou.nl     page 30