Abstract—
In recent years, the boundaries between e-commerce and social networking have become increasingly blurred. Many e-commerce websites support the mechanism of social login where users can sign on the websites using their social network identities such as their Facebook or Twitter accounts. Users can also post their newly purchased products on microblogs with links to the e-commerce product web pages. In this paper, we propose a novel solution for cross-site cold-start product recommendation which aims to recommend products from e-commerce websites to users at social networking sites in “cold-start” situations, a problem which has rarely been explored before. A major challenge is how to leverage knowledge extracted from social networking sites for a cross-site cold-start product recommendation. We propose to use the linked users across social networking sites and e-commerce websites (users who have social networking accounts and have made purchases on e-commerce websites) as a bridge to map users’ social networking features to another feature representation for a product recommendation. In specific, we propose learning both users’ and products’ feature representations (called user embeddings and product embeddings, respectively) from data collected from e-commerce websites using recurrent neural networks and then apply a modified gradient boosting trees method to transform users’ social networking features into user embeddings. We then develop a feature-based matrix factorization approach which can leverage the learned user embeddings for a cold-start product recommendation. Experimental results on a large dataset constructed from the largest Chinese microblogging service SINA WEIBO and the largest Chinese B2C e-commerce website JINGDONG have shown the effectiveness of our proposed framework.
Index Terms—e-commerce, product recommender, product demographic, microblogs, recurrent neural networks.
INTRODUCTION
In recent years, the boundaries between e-commerce and social networking have become increasingly blurred. E-commerce websites such as eBay feature many of the characteristics of social networks, including real-time status updates and interactions between its buyers and sellers. Some e-commerce websites also support the mechanism of social login, which allows new users to sign in with their existing login information from social networking services such as Facebook, Twitter or Google+. Both Facebook and Twitter have introduced a new feature last year that allows users to buy products directly from their websites by clicking a “buy” button to purchase items in adverts or other posts. In China, the e-commerce company Alibaba has made a strategic investment in SINA WEIBO1 where ALIBABA product adverts can be directly delivered to SINA WEIBO users. With the new trend of conducting e-commerce activities on social networking sites, it is important to leverage knowledge extracted from social networking sites for the development of product recommender systems. In this paper, we study an interesting problem of recommending products from e-commerce websites to users at social networking sites who do not have historical purchase records, i.e., in “cold-start” situations.
We called it cross-site cold-start product recommendation. Although online product recommendation has been extensively studied before, most studies only focus on constructing solutions within certain e-commerce websites and mainly utilize users’ historical transaction records. To the best of our knowledge, cross-site cold-start product recommendation has been rarely studied before. In our problem setting here, only the users’ social networking information is available and it is a challenging task to transform the social networking information into latent user features which can be effectively used for a product recommendation. To address this challenge, we propose to use the linked users across social networking sites and e-commerce websites (users who have social networking accounts and have made purchases on e-commerce websites) as a bridge to map users’ social networking features to latent features for a product recommendation. In specific, we propose learning both users’ and products’ feature representations (called user embeddings and product embeddings, respectively) from data collected from e-commerce websites using recurrent neural networks and then apply a modified gradient boosting trees method to transform users’ social networking features into user embeddings. We then develop a feature based matrix factorization approach which can leverage the learned user embeddings for a cold-start product recommendation.
We built our dataset from the largest Chinese microblogging service SINA WEIBO2 and the largest Chinese B2C e-commerce website JINGDONG3, containing a total of 20,638 linked users. The experimental results on the dataset have shown the feasibility and the effectiveness of our proposed framework.
Our major contributions are summarised below:
• We formulate a novel problem of recommending products from an e-commerce website to social networking users in “cold-start” situations. To the best of our knowledge, it has been rarely studied before.
• We propose to apply the recurrent neural networks for learning correlated feature representations for both users and products from data collected from an e-commerce website.
• We propose a modified gradient boosting trees method to transform users’ microblogging attributes to latent feature representation which can be easily incorporated for a product recommendation.
• We propose and instantiate a feature-based matrix factorization approach by incorporating user and product features for a cold-start product recommendation.
In recent years, the boundaries between e-commerce and social networking have become increasingly blurred. Many e-commerce websites support the mechanism of social login where users can sign on the websites using their social network identities such as their Facebook or Twitter accounts. Users can also post their newly purchased products on microblogs with links to the e-commerce product web pages. In this paper, we propose a novel solution for cross-site cold-start product recommendation which aims to recommend products from e-commerce websites to users at social networking sites in “cold-start” situations, a problem which has rarely been explored before. A major challenge is how to leverage knowledge extracted from social networking sites for a cross-site cold-start product recommendation. We propose to use the linked users across social networking sites and e-commerce websites (users who have social networking accounts and have made purchases on e-commerce websites) as a bridge to map users’ social networking features to another feature representation for a product recommendation. In specific, we propose learning both users’ and products’ feature representations (called user embeddings and product embeddings, respectively) from data collected from e-commerce websites using recurrent neural networks and then apply a modified gradient boosting trees method to transform users’ social networking features into user embeddings. We then develop a feature-based matrix factorization approach which can leverage the learned user embeddings for a cold-start product recommendation. Experimental results on a large dataset constructed from the largest Chinese microblogging service SINA WEIBO and the largest Chinese B2C e-commerce website JINGDONG have shown the effectiveness of our proposed framework.
Index Terms—e-commerce, product recommender, product demographic, microblogs, recurrent neural networks.
INTRODUCTION
In recent years, the boundaries between e-commerce and social networking have become increasingly blurred. E-commerce websites such as eBay feature many of the characteristics of social networks, including real-time status updates and interactions between its buyers and sellers. Some e-commerce websites also support the mechanism of social login, which allows new users to sign in with their existing login information from social networking services such as Facebook, Twitter or Google+. Both Facebook and Twitter have introduced a new feature last year that allows users to buy products directly from their websites by clicking a “buy” button to purchase items in adverts or other posts. In China, the e-commerce company Alibaba has made a strategic investment in SINA WEIBO1 where ALIBABA product adverts can be directly delivered to SINA WEIBO users. With the new trend of conducting e-commerce activities on social networking sites, it is important to leverage knowledge extracted from social networking sites for the development of product recommender systems. In this paper, we study an interesting problem of recommending products from e-commerce websites to users at social networking sites who do not have historical purchase records, i.e., in “cold-start” situations.
We called it cross-site cold-start product recommendation. Although online product recommendation has been extensively studied before, most studies only focus on constructing solutions within certain e-commerce websites and mainly utilize users’ historical transaction records. To the best of our knowledge, cross-site cold-start product recommendation has been rarely studied before. In our problem setting here, only the users’ social networking information is available and it is a challenging task to transform the social networking information into latent user features which can be effectively used for a product recommendation. To address this challenge, we propose to use the linked users across social networking sites and e-commerce websites (users who have social networking accounts and have made purchases on e-commerce websites) as a bridge to map users’ social networking features to latent features for a product recommendation. In specific, we propose learning both users’ and products’ feature representations (called user embeddings and product embeddings, respectively) from data collected from e-commerce websites using recurrent neural networks and then apply a modified gradient boosting trees method to transform users’ social networking features into user embeddings. We then develop a feature based matrix factorization approach which can leverage the learned user embeddings for a cold-start product recommendation.
We built our dataset from the largest Chinese microblogging service SINA WEIBO2 and the largest Chinese B2C e-commerce website JINGDONG3, containing a total of 20,638 linked users. The experimental results on the dataset have shown the feasibility and the effectiveness of our proposed framework.
Our major contributions are summarised below:
• We formulate a novel problem of recommending products from an e-commerce website to social networking users in “cold-start” situations. To the best of our knowledge, it has been rarely studied before.
• We propose to apply the recurrent neural networks for learning correlated feature representations for both users and products from data collected from an e-commerce website.
• We propose a modified gradient boosting trees method to transform users’ microblogging attributes to latent feature representation which can be easily incorporated for a product recommendation.
• We propose and instantiate a feature-based matrix factorization approach by incorporating user and product features for a cold-start product recommendation.
Mumbai Academics Is Mumbai’s First Dedicated Professional Training Center For Training With Spoke And Hub Model With Multiple Verticles . The Strong Foundation Of Mumbai Academics Is Laid By Highly Skilled And Trained Professionals, Carrying Mission To Provide Industry Level Input To The Freshers And Highly Skilled And Trained Software Professionals/Other Professional To IT Companies. 
No comments:
Post a Comment