Predict what kind of users might be followed in Tencent Weibo * Project Background
Recently, social network sites have become tremendously popular. Originated in Western countries, it has an ever-growing market in China. Among all Chinese SNS, Tencent Weibo (An imitation of Twitter in China) is one of the most successful, since it owns almost the biggest customer base which is brought by its star product, QQ. However, there is still a crucial problem for the company to consider and to conquer -- how to make use of this remarkable advantage to provide a best customer experience, as well as develop a greater commercial value, in order to take and maintain the leading position in SNS market. * How is Tencent Weibo being played?
“Item” is a specific user (person/organization/group) in Weibo, which can “Tweet” (post something online), and “Retweet”/”@”/”Comment” (all actions) the tweets of other items. Typically, celebrities, famous organizations, or some well-known groups were selected to form the “items set” for recommendation. The recommendation is evaluated as “success” once a user “Follow” the items which are on the recommendation list. * Description of data
The following datasets are owned: 1) User profile data: user_profile.txt (user_ID, the year of birth, the gender, the number of tweet, tag-IDs) 2) Item data: item.txt (item’s category, keyword) 3) User action data: user_action.txt (@ actions between the users in a certain number of recent days) 4) User SNS data: user_sns.txt (each user’s follow history) 5) User key word data: user_key_word.txt(keywords extracted from the tweet/retweet/comment by each user) 6) Training dataset : reg_log_train.txt 7) Testing dataset : reg_log_test.txt * What is going to be done?
Above all, the profile of Weibo users and their behaviors is recorded in the data. Beneath the “Follow” action in Weibo, there are some implicit relationships between all the data. If such rules are found, the risk of overloading information on users will be largely reduced. By capturing users’ behavioral habits, certain items which may appeal to them will be served. A larger commercial value will be developed, and that is exactly what “Micro-blog marketing”, a buzz in nowadays, intends to achieve. * Why the data is interesting?
A large number of the Weibo users are well-educated and perfectly willing to receive new information. What‘s more, Weibo users are significantly curious shown in many cases. Through analyzing the services and data input of Weibo users, their hobbies, preferences and intention could be learned. As a result, the research will be helpful in finding the potential customers, and then will do a favor for precision marketing in the future.