Generative Models for Item Adoptions UsingSocial Correlation

Abstract:

Users face many choices on the Web when it comes to choosing which product to buy, which video to watch, etc. Inmaking adoption decisions, users rely not only on their own preferences, but also on friends. We call the latter social correlationwhich may be caused by the homophily and social influence effects. In this paper, we focus on modeling social correlation on

users item adoptions. Given a user-user social graph and an item-user adoption graph, our research seeks to answer the followingquestions: whether the items adopted by a user correlate to items adopted by her friends, and how to model item adoptions usingsocial correlation. We propose a social correlation framework that considers a social correlation matrix representing the degreesof correlation from every user to the users friends, in addition to a set of latent factors representing topics of interests of individualusers. Based on the framework, we develop two generative models, namely sequential and unified, and the correspondingparameter estimation approaches. From each model, we devise the social correlation only and hybrid methods for predictingmissing adoption links. Experiments on LiveJournal and Epinions data sets show that our proposed models outperform theapproach based on latent factors only (LDA).

Algorithm:

Expectation Maximization Algorithm:

To solve these models, we propose efficientparameter estimation solutions based on Expectation-Maximization that scale with the number of observedlinks.

Existing System:

Existing system shows that

  1. Social correlation was a result of two processes that happenalternatively over a period of time: “homophily”causing users with similar attributes to form sociallinks, and“influence” causing users with social linksto become more similar in attributes.
  1. Combining different social media improvesthe social influence measure but not item adoption.
  1. They only tryto measure influence, but do not incorporate it tomodel item adoption.
  1. They require strongerassumptions, whereby the directions of social edgesare known, and where the influence direction is alreadyknown (e.g., Twitter users re-tweeted postingsby others).
  2. Some prior work focused on how influence propagatedacross a network. Assuming a propagationframework such that an adoption by a user wouldprobabilistically trigger a similar adoption by herfriends but it does not have any rating or comments facilities for the item.

Proposed System:

Inour work, we are concerned only with the existence ofsocial correlation and its use for doption prediction,and not with the underlying causes (homophily vs.influence), which are not always observable from thedata.Ours is a more generalized approach thatallows any friend to be socially dependent on anyfriend. In such cases, the possible number of influencerscan be very large and their method may notscale up.The fundamental assumption here is that every user is correlated with their friends in the sameway. All that matters is the number of friends whohave adopted an item. In contrast, we do not makethe same assumption. In our approach, a user may becorrelated with each friend differently, and may havedifferent self-dependency values.Existing system proposes a social influence matrix to suggest what items a usershould share to maximize their individual influence . Their matrix measures influencebetween users and items while ours measurebetween users and users. Apart from these we have ratings andcomments facilities to value a product where existing system do not have rating and comments facilities.

Modules:

The system is proposed to have the following modules along with functional requirements.

  1. Administrator
  2. Users
  3. End Users
  4. Comments & ratings
  5. Shopping
  6. Virtual gift

Administrator

This is module where major functionalities will takes place. Administrator can be able to view all the details such as products uploaded by the users for advertisement , shopping details, user details and update products.

Update product: Administrator can able to register new product into the system. The users can be able to view those products and purchase.

Users

Users login into the system providing his credentials (User Name and Password).

Major Functionalities of a user are given below:

  1. Log in into the system.
  2. Send and receive mails.
  3. Check their scrapbook which contains the reply for their posted queries.
  4. Post Queries.
  5. Add comments.
  6. User can search a product by ratings and comments.
  7. Purchase items.
  8. View virtual gifts and order the items.

End Users

Major Functionalities of a user are given below:

  1. Add Rating.
  2. Advertise their product.
  3. Send reply for posted queries.
  4. Post feedback.
  5. Create new account.
  6. Search product by rating and comments.
  7. Shopping.

Comments and Ratings

End users can rate products and these ratings can be viewed by all the other users. Comments can be added for an item by users by login into the system and these comments can be viewed by all the other users.

shopping

Shopping plays a major role in this application. Users can do shopping by login into the system providing his credentials (User Name and Password). They can choose a product with the help of ratings and comments or with self dependency. Users have to enter their full details including address for product delivery. After placing an order their invoice will be displayed. These shopping details are maintained by the administrator.

Virtual gift

All the unused item will be retrieved by the administrator and place an attractive rate and qualities and update it. Users can order those items.

Software Requirements:

Technologies : Asp .Net and C#.Net

Database : MS-SQL Server 2005/2008

IDE : Visual Studio 2008

Hardware Requirements:

Processor : Pentium IV

RAM : 1GB