180 likes | 302 Views
User Selectable Interactive Recommendation System In Mobile Environment. Jung-Min Oh NamMee Moon Presented by: Dhanashree Lale. Mobile Environments used widely. Smart phones are popular Digital content is accessible more to people these days Variety of content has increased
E N D
User Selectable Interactive Recommendation System In Mobile Environment Jung-Min Oh NamMee Moon Presented by: Dhanashree Lale
Mobile Environments used widely • Smart phones are popular • Digital content is accessible more to people these days • Variety of content has increased • The web connectivity of the phone to the wireless networks is powerful.
Limitations of digital content for mobile environment • Screen too small • Information retrieval limited by hardware • Causes an information need for users to have a personalized service and relevant information • To answer all these challenges recommendation systems are developed
Recommender System • A user profile is created between explicit as well as implicit forms • Divided into 3 types • Content-based recommendation • Collaborative filtering recommendation • Hybrid recommendation
Proposed user-selectable recommendation system • Users can choose similar groups by themselves • Advantages: • Extends interactivity • Reflects the feature of social networking an user context • Beyond the desktop experience, this approach causes dynamic components of SGs on the user’s social context
User selectable Recommendation System • PG the preference Genre is derived by analyzing the content the user has watched and/or rated already. • SG the similar groups are derived by cross analysis of PG analysis of group based on user’s personal information
System Design and Experiment • MovieLens Dataset used • Website has user data and movie data • The data set modified by adding PG to the user data. • Server : Apache Tomcat 5.5, JSP+XML • Client : iPhone SDK (Xcode, Interface Builder) 3.2.1, iPhone Simulator V3.1, Android SDK(Android 2.1, Platform 2.1, API Level 7), Android DDMS • Generator : JAVA SDK 1.6, Eclipse 3.5.2, My-Sql 5.1, Mac OS X 10.6.3 Snow Leopard, Windows 7
User Similar Group Design 2 step process • Pull out PGs of all users Pull out top 3 genres which have more than 25% rate • Utilize the PG and user’s personal information at the same time
SG similarity calculation • The closer the value of Pearson correlation coefficient is to 1, the higher the similarity is. • The closer to -1, the bigger difference there is with preferences.
Performance Evaluation • MAE is used as a measure. • Best performance : age, gender, occupation together • Worst performance : age, location, occupation together • Good performance when occupation is considered.
Conclusion • In this paper, a flexible and interactive recommendation system in terms of Web 2.0 using collaborative filtering is described. • This paper finds it to be significant that in a mobile environment a user can select and change the interesting group easily to affect their own recommendation. • This proposed system has scalability to ready a future ubiquitous environment. • Further improvements still required to make them effective