370 likes | 404 Views
Learn how recommender systems provide tailored product recommendations to boost online sales and reduce information overload for users. Explore various types and challenges of recommendation systems through the case study of Amazon.
E N D
Recommendation Systems Sunantha Sodsee Information Technology King Mongkut’s University of Technology North Bangkok
E-Commerce • How to enhance E-commerce sales? • Browsers into buyers • Cross-sell Recommender Systems!!
What are recommender systems? • Recommender systems are systems which provide recommendations to a user • Too much information (information overload) • Users have too many choices • Recommend different products for users, suited to their tastes. • Assist users in finding information • Reduce search and navigation time
Case Study: Amazon www.amazon.com
Cross Sell? Rating?
Which Sources of Information? • Sources of information for recommendations: [1] • Browsing and searching data • Purchase data • Feedback provided by the users • Textual comments • Expert recommendations • E-mail • Rating
Type of Recommendations [2] • Population-based • The most popular news articles, or searches, or downloads • Frequently add content • No user tracking needed.
Type of Recommendations [2] • Item-to-item • Content-based • One item is recommended based on the user’s indication that they like another item. • If you like Lord of the Rings, you’ll like Legend.
Type of Recommendations [2] • Challenges with item-to-item: • Getting users to tell you what they like • Financial and time reasons • Getting enough data to make “novel” predictions. • What users really want are recommendations for things they’re not aware of.
Type of Recommendations [2] • Item-to-item • Most effective when you have metadata that lets you automatically relate items. • Genre, actors, director, etc. • Also best when decoupled from payment • Users should have an incentive to rate items truthfully.
Type of Recommendations [2] • User-based • “Users who bought X like Y.” • Each user is represented by a vector indicating his ratings for each product. • Users with a small distance between each other are similar. • Find a similar user and recommend things they like that you haven’t rated.
Type of Recommendations [2] • User-based • Advantages: • Users don’t need to rate much. • No info about products needed. • Easy to implement • Disadvantages • Pushes users “toward the middle” – products with more ratings carry more weight. • How to deal with new products? • Many products and few users -> lots of things don’t get recommended.
Type of Recommendations: General [1] • Content-based Recommender System • Recommend items similar to those users preferred in the past • User profiling is the key • Items/content usually denoted by keywords • Matching “user preferences” with “item characteristics” … works for textual information • Vector Space Model widely used
Type of Recommendations: General [1] • Not all content is well represented by keywords, e.g. images • Items represented by same set of features are indistinguishable • Overspecialization: unrated items not shown • Users with thousands of purchases is a problem • New user: No history available • Shouldn’t show items that are too different, or too similar
Type of Recommendations: General [1] Collaborative Recommender System • Memory-based collaborative filtering techniques • Main problems: scalability and handling of new users • Model-based collaborative filtering techniques • High accuracy of prediction • No need for searching the whole user-item rating matrix (grouping users into models)
Type of Recommendations: General[1] Collaborative Recommender System • Use other users recommendations (ratings) to judge item’s utility • Key is to find users/user groups whose interests match with the current user • Vector Space model widely used (directions of vectors are user specified ratings) • More users, more ratings: better results • Can account for items dissimilar to the ones seen in the past too • Example: Movielens.org
Type of Recommendations: General[1] • Different users might use different scales. Possible solution: weighted ratings, i.e. deviations from average rating • Finding similar users/user groups isn’t very easy • New user: No preferences available • New item: No ratings available • Demographic filtering is required • Multi-criteria ratings is required
Type of Recommendations: Example[1] Cluster Models • Create clusters or groups • Put a customer into a category • Classification simplifies the task of user matching • More scalability and performance • Lesser accuracy than normal collaborative filtering method
Type of Recommendations: Example[1] Item to item collaboration (one that Amazon.com uses) • Compute similarity between item pairs • Combine the similar items into recommendation list • Vector corresponds to an item, and directions correspond to customers who have purchased them • “Similar items” table built offline • Example: Amazon.com
Type of Recommendations: Example[1] Knowledge based RS • Use knowledge of users and items • Conversational Interaction used to establish current user preferences • i.e. “more like this”, “less like that”, “none of those” … • No user profiles maintained, preferences drawn through manual interaction • Query by example … tweaking the source example to fetch results
How RS Work? • Similarity Measurement [4] • For two data objects, X = (x1, x2, . . . , xn) and Y =(y1, y2, . . . , yn), the popular Minkowski distance is defined as • where n is the dimension number of the object and xi, yi are the values of the ith dimension of object X and Y respectively, and q is a positive integer. When q = 1, d is Manhattan distance; when q = 2, d is Euclidian distance
How RS Work? • Similarity wu,vbetween two users u and v, or wi,jbetween two items i and j, is measured by computing the Pearson correlation [4] • where the i ∈ I summations are over the items that both the users u and v have rated and is the average rating of the co-rated items of the u-th user
Prediction and Recommendation Computation • To make a prediction for the active user, a, on a certain item, i, we can take a weighted average of all the ratings on that item according to the following formula[4]
Challenging: # Users and # Items • Clustering Algorithms [5]
References [1] Aalap Kohojkar, Yang Liu, Zhan Shi, “Recommender Systems”, March 31, 2008. [2] Maria Fasli, “Agent Technology for e-Commerce”, http://cswww.essex.ac.uk/staff/mfasli/ATe-Commerce.htm [3]Amit Tiroshi, Tsvi Kuflik, Judy Kay and Bob Kummerfeld,“Recommender Systems and the Social Web”, International Workshop at UMAP2011 on Augmenting User Models with Real World Experiences to Enhance Personalization and Adaptation, July 15, 2011. [4] Xiaoyuan Su, Taghi M. Khoshgoftaar, “A Survey of Collaborative Filtering Techniques”, Advances in Artificial Intelligence, Vol. 2009, 2009. [5]Badrul M. Sarwar, George Karypis , Joseph Konstan, and John Riedl, “Recommender Systems for Large-scale E-Commerce: Scalable Neighborhood Formation Using Clustering”, The Fifth International Conference on Computer and Information Technology (ICCIT 2002) , 2002. [6] Louis E. Frenzel, Jr., “Principles of Electronic Communication Systems”, The third edition, McGraw-Hill, 2008.