160 likes | 258 Views
Behavioral Analysis and Personalization. By Srishti Gahlot (sg2856). What do you mean by online behavior? Why do we need to analyze online behavior and personalize it? How do we analyze this behavior? Comparison between Amazon and ebay Security issues Future work .
E N D
Behavioral Analysis and Personalization By SrishtiGahlot (sg2856)
What do you mean by online behavior? Why do we need to analyze online behavior and personalize it? How do we analyze this behavior? Comparison between Amazon and ebay Security issues Future work
What do you mean by online behavior? Online behavior refers to organized and unorganized interactions with both human and nonhuman elements in online environments. Salient features of online behavior: Sociability Utility Reciprocity
Why do we need to analyze online behavior and personalize it? Expanding WWW and dependence on web Ecommerce and marketing firm can Campaign and advertize Customize shopping experience Profit
How do we analyze this behavior? Recommendation system: analyzes patterns of user interest in products to provide personalized recommendations that suit a user’s taste. Eg Amazon, Netflix Types of strategies used in Recommendation Systems: a. Content based filtering b. Collaborative filtering
Content based filtering Creates profile (demographic info, questionnaire) Gather external information Independent of other users
Collaborative filtering filtering for information or patterns using collaboration analyzes relationships between users and interdependencies among products to identify new user-item associations. Main Aim 1. Look for users who share the same rating patterns with the active user 2. Use the ratings from those like-minded users to calculate a prediction for the active user
Architecture N dimensional vector of items Components are positive for purchased or positively rated items Multiplies with inverse frequency Sparse vector Recommendation is based on similarity Cosine similarity
Recommendation System for Amazon Uses item-to-item collaborative filtering Purchased and rated itemsrecommendation list.
Recommendation System for eBay Content based filtering recommendation Feedback Profile Dissatisfied/neutral/satisfied + comments
Security Issues Risk of unwanted exposure of information The recommender can violate the users trust in three ways: Exposure: Undesired access to personal user information Bias: Manipulation of users’ recommendations to inappropriately change the items that are recommended
Sabotage: Intentionally reducing the recommendation accuracy of a recommender A recommendation system must win the trust of its users on two following grounds: The system will protect their information appropriately The recommendation made by the system should be accurate
Future work Experiments are being done to somehow combine these two algorithms to come up with a more powerful recommender system Content-Boosted Collaborative Filtering algorithm(CBCF) performance is not very large (4%)
Thank you Question or Comments