Presentation is loading. Please wait.

Presentation is loading. Please wait.

- 1 -. - 2 - Hybrid recommender systems Hybrid: combinations of various inputs and/or composition of different mechanism Knowledge-based: "Tell me what.

Similar presentations

Presentation on theme: "- 1 -. - 2 - Hybrid recommender systems Hybrid: combinations of various inputs and/or composition of different mechanism Knowledge-based: "Tell me what."— Presentation transcript:

1 - 1 -

2 - 2 - Hybrid recommender systems Hybrid: combinations of various inputs and/or composition of different mechanism Knowledge-based: "Tell me what fits based on my needs" Content-based: "Show me more of the same what I've liked " Collaborative: "Tell me what's popular among my peers"

3 - 3 - Hybrid recommender systems  Idea of crossing two (or more) species/implementations –hybrida [lat.]: denotes an object made by combining two different elements –Overcome some of the shortcomings (i.e., cold-start problem) –Reach desirable properties that is not present in parent individuals  Different hybridization designs –Parallel use of several systems –Monolithic exploiting different features –Pipelined invocation of different systems

4 - 4 - Monolithic hybridization design  Consist of only single recommendation component  Integrates multiple recommendation algorithms and combines multiple feature/knowledge sources (as input data).  Monolithic recommendation component can be designed based on two strategies: –feature combination hybrid –feature augmentation hybrid

5 - 5 - Monolithic hybridization designs: Feature combination Example C Community and product knowledge Feature combination strategy Social/collaborative features: movies liked by user Content features: movie1 is Comedy, movie2 is drama Hybrid features: user likes many movies that are comedies, …

6 - 6 - Monolithic hybridization designs: Feature combination (cont.)  Predict similar users to Alice based on –Pure collaborative approach  User1 and User2 are similar to Alice –Feature combination approach  User2 behaves in very indeterminate manners  User1 and Alice are similar. Hybrid input features

7 - 7 - Monolithic hybridization designs: Feature augmentation  In feature augmentation, we seek to have the contributing recommender produce some features that augment the knowledge source used by the primary(actual) recommender.  e.g. content-boosted collaborative filtering [Prem Melville et al. 2002] –Content-Based Contributing / Collaborative Actual 1.The content-based algorithm is trained on the user profile (e.g. Alice) e.g. which items would user like/dislike 2.The collaborative algorithm retrieves e.g. candidate restaurants from similar users (to Alice)  e.g. User1 likes Macdonald, User2 likes pizza hut 3. These candidates (restaurant ) will be rated by the content-based classifier and those ratings are used to augment the user profile (e.g. Alice ); thus filling it out with more ratings.  e.g. content-base classifier predict Alice would like pizza hut and would dislike Macdonald 4.The collaborative recommendation process is performed again with a new augmented profile (e.g. measure the similarity of users again).  Similarity degree of Alice and User1 and Alice and User2 would change with the new feature augmented profile

8 - 8 - Parallelized hybridization design  Output of several existing implementations combined to generate final recommendation  Three hybridization strategies: –Mixed hybrids: the output of all recommenders presented side by side to a user –Weighted hybrids: Weights can be learned dynamically –Switching hybrids: Extreme case of dynamic weighting is switching

9 - 9 - Parallelized hybridization design: Weighted  A weighted hybridization strategy combines the recommendations of two or more recommendation systems by computing weighted sum of their scores.  Given n different recommenders rec k and their associated weights, the final recommendation score of item i for user u is: where

10 - 10 - Recommender weighted(0.5:0.5) Item10.651 Item20.452 Item30.353 Item40.054 Item50.00 Parallelized hybridization design: Weighted Recommender 1 Item10.51 Item20 Item30.32 Item40.13 Item50 Recommender 2 Item10.82 Item20.91 Item30.43 Item40 Item50

11 - 11 - Parallelized hybridization design: Weighted  BUT, how to derive weights? –Estimate, e.g. by empirical bootstrapping –Dynamic adjustment of weights  Empirical bootstrapping –Historic data is needed –Compute different weightings –Decide which one does best  Dynamic adjustment of weights –Start with for instance uniform weight distribution –For each user adopt weights to minimize error of prediction

12 - 12 - Parallelized hybridization design: Dynamic weighting  Let's assume Alice actually bought items 1 and 4: (r 1 =1,r 4 = 1) –Identify weighting that minimizes Mean Absolute Error (MAE) Absolute errors and MAE Beta1Beta2rec1rec2errorMAE 0.10.9Item10. Item40.10.00.99 0.30.7Item10. Item40.10.00.97 0.5 Item10.50.80.350.65 Item40.10.00.95 0.70.3Item10.50.80.410.67 Item40.10.00.93 0.90.1Item10.50.80.470.69 Item40.10.00.91  MAE improves as rec2 is weighted more strongly

13 - 13 - Parallelized hybridization design: Dynamic weighting  BUT: didn't rec1 actually rank Item 1 and item 4 higher?  Be careful when weighting! –Recommenders need to assign comparable scores over all users and items  scores generated by different recommenders for the same items should not be high- variance  Some score transformation could be necessary –Stable weights require several user ratings Recommender 1 Item10.51 Item20 Item30.32 Item40.13 Item50 Recommender 2 Item10.82 Item20.91 Item30.43 Item40 Item50

14 - 14 - Parallelized hybridization design: Switching  Requires a mediator that decides which recommender should be used in a specific situation, depending on user profiles and the quality of recommendation results  Special case of dynamic weights (all except one Beta is 0)  Example: –To overcome the cold-start problem for new user  Use knowledge-based first, when users’ ratings collected, use collaborative filtering –To overcome the cold-start problem for new item  Use content-based filtering first, then use collaborative filtering when item’s ratings collected

15 - 15 - Parallelized hybridization design: Mixed  Combines the results of different recommender systems at the level of user interface  Results of different techniques are presented together  Recommendation result for user and item is the set of tuples for each of its constituting recommenders rec k

16 - 16 - Pipelined hybridization designs  In pipelined hybridization several recommenders sequentially build on each other before the final one produces recommendation for the user  Pipelined hybrids differs according to the type of output they generate for the next recommender –Cascade : the preceding component deliver a recommendation list to subsequent component –Meta-level: the preceding component preprocess input data to build a model for subsequent component

17 - 17 - Pipelined hybridization designs: Cascade  Refinement of recommendation lists  Successor's recommendations are restricted by predecessor  Where for all k > 1  Subsequent recommender may not introduce additional items  Thus produces very precise results

18 - 18 - Pipelined hybridization designs: Cascade  Recommendation list is continually reduced  There is a chance that no recommendation be made –Cascade strategy should be combined by switching strategy to generate enough recommendation  First recommender can only make a recommendation list –e.g. knowledge-based recommender produce lists of unsorted items  Subsequent (second) recommender assigns score or exclude an item –Ordering and refinement (e.g. collaborative)

19 - 19 - Pipelined hybridization designs: Cascade Recommender 1 Item10.51 Item20 Item30.32 Item40.13 Item50 Recommender 2 Item10.81 Item20 Item30.42 Item40 Item50 Recommender 3 Item10.702 Item20.00 Item30.851 Item40.00 Item50.00 Recommendation list Ordering and refinement

20 - 20 - Pipelined hybridization designs: Meta-level  Successor exploits a model delta built by predecessor – one recommender builds a (user) model that is exploited by the principal recommender to make a prediction  Examples : –Fab system [Balabanovic and Shham,1997]:  Based on meta-level pipeline hybridization design  Online news domain  CB recommender builds user models based on weighted term vectors  CF identifies similar peers based on these user models but makes recommendations based on users’ ratings

21 - 21 - Limitations of hybridization strategies  Only few works exist that compare different recommendation strategies, specially their hybrids –Like for instance, [Robin Burke 2002] –Most datasets do not have all the requirements to compare different recommendation paradigms  i.e. ratings, requirements, item features, domain knowledge, critiques rarely available in a single dataset –Thus few conclusions that are supported by empirical findings  Monolithic: some preprocessing steps or minor modification in the recommender algorithms and data structure are required– to include more knowledge on the feature level  Parallel: requires careful matching and weighting of scores from different predictors  Pipelined: works well for two antithetic (mutually incompatible) approaches  Netflix competition – "stacking" recommender systems –Parallel hybrid design with adaptive weights/switching for recommenders –Weighted design based on >100 predictors – recommendation functions –Adaptive switching of weights based on user model, context and meta-features

22 - 22 - Literature  [Robin Burke 2002] Hybrid recommender systems: Survey and experiments, User Modeling and User- Adapted Interaction 12 (2002), no. 4, 331-370.  [Prem Melville et al. 2002] Content-Boosted Collaborative Filtering for Improved Recommendations, Proceedings of the 18th National Conference on Artificial Intelligence (AAAI) (Edmonton,CAN), American Association for Artificial Intelligence, 2002, pp. 187-192.  [Roberto Torres et al. 2004] Enhancing digital libraries with techlens, International Joint Conference on Digital Libraries (JCDL'04) (Tucson, AZ), 2004, pp. 228-236.  [Chumki Basuet al. 1998] Recommendation as classification: using social and content-based information in recommendation, In Proceedings of the 15th National Conference on Artificial Intelligence (AAAI'98) (Madison, Wisconsin, USA States), American Association for Artificial Intelligence, 1998, pp. 714-720.

Download ppt "- 1 -. - 2 - Hybrid recommender systems Hybrid: combinations of various inputs and/or composition of different mechanism Knowledge-based: "Tell me what."

Similar presentations

Ads by Google